 TensorFlow.org에서 보기 |
 Run in Google Colab |
 View source on GitHub |
개요
그래프 정규화는 Neural Graph Learning의 더 넓은 패러다임에서 사용되는 특정 기술입니다(Bui et al., 2018). 핵심 아이디어는 레이블이 지정된 데이터와 레이블이 없는 데이터를 모두 활용하여 그래프 정규화 목표를 갖고 신경망 모델을 훈련하는 것입니다.
이 튜토리얼에서는 그래프 정규화를 사용하여 자연(유기적) 그래프를 형성하는 문서를 분류하는 방법을 살펴봅니다.
Neural Structured Learning(NSL) 프레임워크를 사용하여 그래프 정규화 모델을 생성하는 일반적인 방법은 다음과 같습니다.
- 입력 그래프 및 샘플 특성에서 훈련 데이터를 생성합니다. 그래프의 노드는 샘플에 해당하고, 그래프의 간선은 샘플 쌍 간의 유사성에 해당합니다. 결과 훈련 데이터에는 원래 노드 특성 외에도 이웃 특성이 포함됩니다.
Keras순차, 함수형 또는 서브 클래스 API를 사용하여 신경망을 기본 모델로 만듭니다.- NSL 프레임워크에서 제공하는
GraphRegularization래퍼 클래스로 기본 모델을 래핑하여 새 그래프Keras모델을 만듭니다. 이 새로운 모델은 훈련 목표에서 그래프 정규화 손실을 정규화 항으로 포함합니다. - 그래프
Keras모델을 훈련하고 평가합니다.
설정
Neural Structured Learning 패키지를 설치합니다.
pip install --quiet neural-structured-learning종속성 및 가져오기
import neural_structured_learning as nsl
import tensorflow as tf
# Resets notebook state
tf.keras.backend.clear_session()
print("Version: ", tf.__version__)
print("Eager mode: ", tf.executing_eagerly())
print(
"GPU is",
"available" if tf.config.list_physical_devices("GPU") else "NOT AVAILABLE")
Version: 2.4.0 Eager mode: True GPU is available
Cora 데이터세트
Cora 데이터세트는 노드가 머신러닝 논문을 나타내고 간선이 논문 쌍 간의 인용을 나타내는 인용 그래프입니다. 관련된 작업은 각 논문을 7가지 범주 중 하나로 분류하는 것을 목표로 하는 문서 분류입니다. 즉, 7개의 클래스가 있는 다중 클래스 분류 문제입니다.
그래프
원래 그래프에는 방향이 있습니다. 그러나 이 예에서는 이 그래프의 방향 없는 버전을 고려합니다. 따라서 A 논문이 B 논문을 인용하면 B 논문도 A를 인용한 것으로 간주합니다. 이것이 반드시 사실은 아니지만, 이 예에서는 인용을 유사성에 대한 프록시로 간주하며, 일반적으로 교환 속성입니다.
특성
입력의 각 논문에는 효과적으로 두 가지 특성이 포함되어 있습니다.
Words: 종이에 있는 텍스트를 표현한 밀집 멀티-핫 단어 주머니(bag-of-words)입니다. Cora 데이터세트의 어휘에는 1433개의 고유한 단어가 포함되어 있습니다. 따라서 이 특성의 길이는 1433이고, 위치 'i'의 값은 주어진 논문에서 해당 어휘의 단어 'i'가 존재하는지 여부를 나타내는 0/1입니다.
Label: 논문의 클래스 ID(카테고리)를 나타내는 단일 정수입니다.
Cora 데이터세트 다운로드하기
wget --quiet -P /tmp https://linqs-data.soe.ucsc.edu/public/lbc/cora.tgztar -C /tmp -xvzf /tmp/cora.tgz
cora/ cora/README cora/cora.cites cora/cora.content
Cora 데이터를 NSL 형식으로 변환하기
Cora 데이터세트를 전처리하고 Neural Structured Learning에 필요한 형식으로 변환하기 위해 NSL github 리포지토리에 포함된 'preprocess_cora_dataset.py' 스크립트를 실행합니다. 이 스크립트는 다음을 수행합니다.
- 원래 노드 특성과 그래프를 사용하여 이웃 특성을 생성합니다.
tf.train.Example인스턴스를 포함하는 훈련 및 테스트 데이터 분할을 생성합니다.- 결과 훈련 및 테스트 데이터를
TFRecord형식으로 유지합니다.
!wget https://raw.githubusercontent.com/tensorflow/neural-structured-learning/master/neural_structured_learning/examples/preprocess/cora/preprocess_cora_dataset.py
!python preprocess_cora_dataset.py \
--input_cora_content=/tmp/cora/cora.content \
--input_cora_graph=/tmp/cora/cora.cites \
--max_nbrs=5 \
--output_train_data=/tmp/cora/train_merged_examples.tfr \
--output_test_data=/tmp/cora/test_examples.tfr
--2021-01-15 02:26:25-- https://raw.githubusercontent.com/tensorflow/neural-structured-learning/master/neural_structured_learning/examples/preprocess/cora/preprocess_cora_dataset.py
Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.192.133, 151.101.128.133, 151.101.64.133, ...
Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.192.133|:443... connected.
HTTP request sent, awaiting response...
200 OK
Length: 11640 (11K) [text/plain]
Saving to: ‘preprocess_cora_dataset.py’
preproces 0%[ ] 0 --.-KB/s
preprocess_cora_dat 100%[===================>] 11.37K --.-KB/s in 0.001s
2021-01-15 02:26:26 (20.7 MB/s) - ‘preprocess_cora_dataset.py’ saved [11640/11640]
2021-01-15 02:26:26.653369: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
Reading graph file: /tmp/cora/cora.cites...
Done reading 5429 edges from: /tmp/cora/cora.cites (0.01 seconds).
Making all edges bi-directional...
Done (0.01 seconds). Total graph nodes: 2708
Joining seed and neighbor tf.train.Examples with graph edges...
Done creating and writing 2155 merged tf.train.Examples (1.51 seconds).
Out-degree histogram: [(1, 386), (2, 468), (3, 452), (4, 309), (5, 540)]
Output training data written to TFRecord file: /tmp/cora/train_merged_examples.tfr.
Output test data written to TFRecord file: /tmp/cora/test_examples.tfr.
Total running time: 0.05 minutes.
전역 변수
훈련 및 테스트 데이터에 대한 파일 경로는 위의 'preprocess_cora_dataset.py' 스크립트를 호출하는 데 사용된 명령 줄 플래그 값을 기반으로 합니다.
### Experiment dataset
TRAIN_DATA_PATH = '/tmp/cora/train_merged_examples.tfr'
TEST_DATA_PATH = '/tmp/cora/test_examples.tfr'
### Constants used to identify neighbor features in the input.
NBR_FEATURE_PREFIX = 'NL_nbr_'
NBR_WEIGHT_SUFFIX = '_weight'
하이퍼 매개변수
HParams의 인스턴스를 사용하여 훈련 및 평가에 사용되는 다양한 하이퍼 매개변수 및 상수를 포함합니다. 아래에서 각각에 대해 간략하게 설명합니다.
num_classes: 총 7개의 클래스가 있습니다.
max_seq_length: 어휘의 크기이며, 입력의 모든 인스턴스는 밀집 멀티-핫, 단어 주머니(bag-of-words)의 표현을 갖습니다. 즉, 단어의 값이 1이면 해당 단어가 입력에 있음을 나타내고, 값이 0이면 그렇지 않음을 나타냅니다.
distance_type: 샘플을 이웃으로 정규화하는 데 사용되는 거리 메트릭입니다.
graph_regularization_multiplier: 전체 손실 함수에서 그래프 정규화 항의 상대적 가중치를 제어합니다.
num_neighbors: 그래프 정규화에 사용되는 이웃의 수입니다. 이 값은
preprocess_cora_dataset.py를 실행할 때 위에 사용된max_nbrs명령 줄 인수보다 작거나 같아야 합니다.num_fc_units: 신경망에서 완전 연결된 레이어의 수입니다.
train_epochs: 훈련 epoch의 수입니다.
batch_size: 훈련 및 평가에 사용되는 배치 크기입니다.
dropout_rate: 각 완전 연결 레이어의 드롭아웃 비율을 제어합니다.
eval_steps: 평가가 완료된 것으로 간주하기 전에 처리할 배치의 수입니다.
None으로 설정하면, 테스트세트의 모든 인스턴스가 평가됩니다.
class HParams(object):
"""Hyperparameters used for training."""
def __init__(self):
### dataset parameters
self.num_classes = 7
self.max_seq_length = 1433
### neural graph learning parameters
self.distance_type = nsl.configs.DistanceType.L2
self.graph_regularization_multiplier = 0.1
self.num_neighbors = 1
### model architecture
self.num_fc_units = [50, 50]
### training parameters
self.train_epochs = 100
self.batch_size = 128
self.dropout_rate = 0.5
### eval parameters
self.eval_steps = None # All instances in the test set are evaluated.
HPARAMS = HParams()
훈련 및 테스트 데이터 로드하기
이 노트북의 앞부분에서 설명한 것처럼 입력 훈련 및 테스트 데이터는 'preprocess_cora_dataset.py'에 의해 생성되었습니다. 데이터를 두 개의 tf.data.Dataset 객체로 로드합니다. 하나는 훈련용이고 다른 하나는 테스트용입니다.
모델의 입력 레이어에서 각 샘플의 'words' 및 'label' 특성뿐만 아니라 hparams.num_neighbors 값을 기반으로 해당 이웃 특성도 추출합니다. 이웃이 hparams.num_neighbors보다 적은 인스턴스에는 존재하지 않는 이웃 특성에 대해 더미 값이 할당됩니다.
def make_dataset(file_path, training=False):
"""Creates a `tf.data.TFRecordDataset`.
Args:
file_path: Name of the file in the `.tfrecord` format containing
`tf.train.Example` objects.
training: Boolean indicating if we are in training mode.
Returns:
An instance of `tf.data.TFRecordDataset` containing the `tf.train.Example`
objects.
"""
def parse_example(example_proto):
"""Extracts relevant fields from the `example_proto`.
Args:
example_proto: An instance of `tf.train.Example`.
Returns:
A pair whose first value is a dictionary containing relevant features
and whose second value contains the ground truth label.
"""
# The 'words' feature is a multi-hot, bag-of-words representation of the
# original raw text. A default value is required for examples that don't
# have the feature.
feature_spec = {
'words':
tf.io.FixedLenFeature([HPARAMS.max_seq_length],
tf.int64,
default_value=tf.constant(
0,
dtype=tf.int64,
shape=[HPARAMS.max_seq_length])),
'label':
tf.io.FixedLenFeature((), tf.int64, default_value=-1),
}
# We also extract corresponding neighbor features in a similar manner to
# the features above during training.
if training:
for i in range(HPARAMS.num_neighbors):
nbr_feature_key = '{}{}_{}'.format(NBR_FEATURE_PREFIX, i, 'words')
nbr_weight_key = '{}{}{}'.format(NBR_FEATURE_PREFIX, i,
NBR_WEIGHT_SUFFIX)
feature_spec[nbr_feature_key] = tf.io.FixedLenFeature(
[HPARAMS.max_seq_length],
tf.int64,
default_value=tf.constant(
0, dtype=tf.int64, shape=[HPARAMS.max_seq_length]))
# We assign a default value of 0.0 for the neighbor weight so that
# graph regularization is done on samples based on their exact number
# of neighbors. In other words, non-existent neighbors are discounted.
feature_spec[nbr_weight_key] = tf.io.FixedLenFeature(
[1], tf.float32, default_value=tf.constant([0.0]))
features = tf.io.parse_single_example(example_proto, feature_spec)
label = features.pop('label')
return features, label
dataset = tf.data.TFRecordDataset([file_path])
if training:
dataset = dataset.shuffle(10000)
dataset = dataset.map(parse_example)
dataset = dataset.batch(HPARAMS.batch_size)
return dataset
train_dataset = make_dataset(TRAIN_DATA_PATH, training=True)
test_dataset = make_dataset(TEST_DATA_PATH)
내용을 보기 위해 훈련 데이터세트를 살펴보겠습니다.
for feature_batch, label_batch in train_dataset.take(1):
print('Feature list:', list(feature_batch.keys()))
print('Batch of inputs:', feature_batch['words'])
nbr_feature_key = '{}{}_{}'.format(NBR_FEATURE_PREFIX, 0, 'words')
nbr_weight_key = '{}{}{}'.format(NBR_FEATURE_PREFIX, 0, NBR_WEIGHT_SUFFIX)
print('Batch of neighbor inputs:', feature_batch[nbr_feature_key])
print('Batch of neighbor weights:',
tf.reshape(feature_batch[nbr_weight_key], [-1]))
print('Batch of labels:', label_batch)
Feature list: ['NL_nbr_0_weight', 'NL_nbr_0_words', 'words'] Batch of inputs: tf.Tensor( [[0 0 1 ... 0 0 0] [1 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] ... [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0]], shape=(128, 1433), dtype=int64) Batch of neighbor inputs: tf.Tensor( [[0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] ... [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0]], shape=(128, 1433), dtype=int64) Batch of neighbor weights: tf.Tensor( [1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.], shape=(128,), dtype=float32) Batch of labels: tf.Tensor( [2 2 2 0 0 6 6 1 2 6 2 0 5 6 1 2 5 3 3 0 3 3 2 6 6 2 1 2 1 3 6 4 6 3 0 2 2 1 3 3 6 6 3 2 2 1 2 2 6 6 5 0 6 2 0 2 6 6 2 2 2 5 2 3 3 0 3 3 6 3 6 3 1 2 2 3 3 3 2 3 0 1 2 2 0 2 3 3 2 6 3 2 3 1 2 4 2 1 2 2 3 6 1 2 3 2 5 2 2 3 2 2 1 1 3 2 1 4 0 2 3 5 2 1 2 2 0 1], shape=(128,), dtype=int64)
내용을 보기 위해 테스트 데이터세트를 살펴보겠습니다.
for feature_batch, label_batch in test_dataset.take(1):
print('Feature list:', list(feature_batch.keys()))
print('Batch of inputs:', feature_batch['words'])
print('Batch of labels:', label_batch)
Feature list: ['words'] Batch of inputs: tf.Tensor( [[0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] ... [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0] [0 0 0 ... 0 0 0]], shape=(128, 1433), dtype=int64) Batch of labels: tf.Tensor( [5 2 2 2 1 2 6 3 2 3 6 1 3 6 4 4 2 3 3 0 2 0 5 2 1 0 6 3 6 4 2 2 3 0 4 2 2 2 2 3 2 2 2 0 2 2 2 2 4 2 3 4 0 2 6 2 1 4 2 0 0 1 4 2 6 0 5 2 2 3 2 5 2 5 2 3 2 2 2 2 2 6 6 3 2 4 2 6 3 2 2 6 2 4 2 2 1 3 4 6 0 0 2 4 2 1 3 6 6 2 6 6 6 1 4 6 4 3 6 6 0 0 2 6 2 4 0 0], shape=(128,), dtype=int64)
모델 정의
그래프 정규화를 사용하는 방법을 보여주기 위해 먼저 이 문제에 대한 기본 모델을 빌드합니다. 2개의 숨겨진 레이어와 그 사이에 드롭아웃이 있는 간단한 피드 포워드 신경망을 사용합니다. tf.Keras 프레임워크에서 지원하는 모든 모델 유형(순차, 함수형 및 서브 클래스)을 사용하여 기본 모델을 생성하는 방법을 설명합니다.
순차 기본 모델
def make_mlp_sequential_model(hparams):
"""Creates a sequential multi-layer perceptron model."""
model = tf.keras.Sequential()
model.add(
tf.keras.layers.InputLayer(
input_shape=(hparams.max_seq_length,), name='words'))
# Input is already one-hot encoded in the integer format. We cast it to
# floating point format here.
model.add(
tf.keras.layers.Lambda(lambda x: tf.keras.backend.cast(x, tf.float32)))
for num_units in hparams.num_fc_units:
model.add(tf.keras.layers.Dense(num_units, activation='relu'))
# For sequential models, by default, Keras ensures that the 'dropout' layer
# is invoked only during training.
model.add(tf.keras.layers.Dropout(hparams.dropout_rate))
model.add(tf.keras.layers.Dense(hparams.num_classes, activation='softmax'))
return model
함수형 기본 모델
def make_mlp_functional_model(hparams):
"""Creates a functional API-based multi-layer perceptron model."""
inputs = tf.keras.Input(
shape=(hparams.max_seq_length,), dtype='int64', name='words')
# Input is already one-hot encoded in the integer format. We cast it to
# floating point format here.
cur_layer = tf.keras.layers.Lambda(
lambda x: tf.keras.backend.cast(x, tf.float32))(
inputs)
for num_units in hparams.num_fc_units:
cur_layer = tf.keras.layers.Dense(num_units, activation='relu')(cur_layer)
# For functional models, by default, Keras ensures that the 'dropout' layer
# is invoked only during training.
cur_layer = tf.keras.layers.Dropout(hparams.dropout_rate)(cur_layer)
outputs = tf.keras.layers.Dense(
hparams.num_classes, activation='softmax')(
cur_layer)
model = tf.keras.Model(inputs, outputs=outputs)
return model
서브 클래스 기본 모델
def make_mlp_subclass_model(hparams):
"""Creates a multi-layer perceptron subclass model in Keras."""
class MLP(tf.keras.Model):
"""Subclass model defining a multi-layer perceptron."""
def __init__(self):
super(MLP, self).__init__()
# Input is already one-hot encoded in the integer format. We create a
# layer to cast it to floating point format here.
self.cast_to_float_layer = tf.keras.layers.Lambda(
lambda x: tf.keras.backend.cast(x, tf.float32))
self.dense_layers = [
tf.keras.layers.Dense(num_units, activation='relu')
for num_units in hparams.num_fc_units
]
self.dropout_layer = tf.keras.layers.Dropout(hparams.dropout_rate)
self.output_layer = tf.keras.layers.Dense(
hparams.num_classes, activation='softmax')
def call(self, inputs, training=False):
cur_layer = self.cast_to_float_layer(inputs['words'])
for dense_layer in self.dense_layers:
cur_layer = dense_layer(cur_layer)
cur_layer = self.dropout_layer(cur_layer, training=training)
outputs = self.output_layer(cur_layer)
return outputs
return MLP()
기본 모델 생성하기
# Create a base MLP model using the functional API.
# Alternatively, you can also create a sequential or subclass base model using
# the make_mlp_sequential_model() or make_mlp_subclass_model() functions
# respectively, defined above. Note that if a subclass model is used, its
# summary cannot be generated until it is built.
base_model_tag, base_model = 'FUNCTIONAL', make_mlp_functional_model(HPARAMS)
base_model.summary()
Model: "model" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= words (InputLayer) [(None, 1433)] 0 _________________________________________________________________ lambda (Lambda) (None, 1433) 0 _________________________________________________________________ dense (Dense) (None, 50) 71700 _________________________________________________________________ dropout (Dropout) (None, 50) 0 _________________________________________________________________ dense_1 (Dense) (None, 50) 2550 _________________________________________________________________ dropout_1 (Dropout) (None, 50) 0 _________________________________________________________________ dense_2 (Dense) (None, 7) 357 ================================================================= Total params: 74,607 Trainable params: 74,607 Non-trainable params: 0 _________________________________________________________________
기본 MLP 모델 훈련하기
# Compile and train the base MLP model
base_model.compile(
optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
base_model.fit(train_dataset, epochs=HPARAMS.train_epochs, verbose=1)
Epoch 1/100 /tmpfs/src/tf_docs_env/lib/python3.6/site-packages/tensorflow/python/keras/engine/functional.py:595: UserWarning: Input dict contained keys ['NL_nbr_0_weight', 'NL_nbr_0_words'] which did not match any model input. They will be ignored by the model. [n for n in tensors.keys() if n not in ref_input_names]) 1/Unknown - 1s 846ms/step - loss: 1.9927 - accuracy: 0.1250 7/Unknown - 1s 11ms/step - loss: 1.9846 - accuracy: 0.1383 12/Unknown - 1s 11ms/step - loss: 1.9795 - accuracy: 0.1369 16/Unknown - 1s 11ms/step - loss: 1.9743 - accuracy: 0.1409 17/17 [==============================] - 1s 12ms/step - loss: 1.9717 - accuracy: 0.1436 Epoch 2/100 1/17 [>.............................] - ETA: 0s - loss: 1.9281 - accuracy: 0.1797 6/17 [=========>....................] - ETA: 0s - loss: 1.8872 - accuracy: 0.2583 11/17 [==================>...........] - ETA: 0s - loss: 1.8791 - accuracy: 0.2745 16/17 [===========================>..] - ETA: 0s - loss: 1.8732 - accuracy: 0.2818 17/17 [==============================] - 0s 12ms/step - loss: 1.8710 - accuracy: 0.2844 Epoch 3/100 1/17 [>.............................] - ETA: 0s - loss: 1.7776 - accuracy: 0.3672 6/17 [=========>....................] - ETA: 0s - loss: 1.7710 - accuracy: 0.3582 11/17 [==================>...........] - ETA: 0s - loss: 1.7740 - accuracy: 0.3466 16/17 [===========================>..] - ETA: 0s - loss: 1.7696 - accuracy: 0.3442 17/17 [==============================] - 0s 11ms/step - loss: 1.7676 - accuracy: 0.3435 Epoch 4/100 1/17 [>.............................] - ETA: 0s - loss: 1.7099 - accuracy: 0.3516 6/17 [=========>....................] - ETA: 0s - loss: 1.6841 - accuracy: 0.3557 11/17 [==================>...........] - ETA: 0s - loss: 1.6815 - accuracy: 0.3547 16/17 [===========================>..] - ETA: 0s - loss: 1.6777 - accuracy: 0.3545 17/17 [==============================] - 0s 12ms/step - loss: 1.6748 - accuracy: 0.3555 Epoch 5/100 1/17 [>.............................] - ETA: 0s - loss: 1.6721 - accuracy: 0.3359 6/17 [=========>....................] - ETA: 0s - loss: 1.6218 - accuracy: 0.3563 11/17 [==================>...........] - ETA: 0s - loss: 1.5979 - accuracy: 0.3762 16/17 [===========================>..] - ETA: 0s - loss: 1.5818 - accuracy: 0.3887 17/17 [==============================] - 0s 11ms/step - loss: 1.5762 - accuracy: 0.3930 Epoch 6/100 1/17 [>.............................] - ETA: 0s - loss: 1.4095 - accuracy: 0.5781 6/17 [=========>....................] - ETA: 0s - loss: 1.4340 - accuracy: 0.5127 11/17 [==================>...........] - ETA: 0s - loss: 1.4199 - accuracy: 0.5140 16/17 [===========================>..] - ETA: 0s - loss: 1.4125 - accuracy: 0.5143 17/17 [==============================] - 0s 12ms/step - loss: 1.4101 - accuracy: 0.5144 Epoch 7/100 1/17 [>.............................] - ETA: 0s - loss: 1.2778 - accuracy: 0.5938 6/17 [=========>....................] - ETA: 0s - loss: 1.2679 - accuracy: 0.5795 11/17 [==================>...........] - ETA: 0s - loss: 1.2500 - accuracy: 0.5883 16/17 [===========================>..] - ETA: 0s - loss: 1.2405 - accuracy: 0.5920 17/17 [==============================] - 0s 12ms/step - loss: 1.2371 - accuracy: 0.5929 Epoch 8/100 1/17 [>.............................] - ETA: 0s - loss: 1.1326 - accuracy: 0.6406 6/17 [=========>....................] - ETA: 0s - loss: 1.1610 - accuracy: 0.5966 11/17 [==================>...........] - ETA: 0s - loss: 1.1582 - accuracy: 0.5940 16/17 [===========================>..] - ETA: 0s - loss: 1.1442 - accuracy: 0.5995 17/17 [==============================] - 0s 11ms/step - loss: 1.1377 - accuracy: 0.6022 Epoch 9/100 1/17 [>.............................] - ETA: 0s - loss: 1.0950 - accuracy: 0.6484 6/17 [=========>....................] - ETA: 0s - loss: 1.0106 - accuracy: 0.6713 10/17 [================>.............] - ETA: 0s - loss: 1.0004 - accuracy: 0.6703 15/17 [=========================>....] - ETA: 0s - loss: 0.9884 - accuracy: 0.6735 17/17 [==============================] - 0s 12ms/step - loss: 0.9814 - accuracy: 0.6762 Epoch 10/100 1/17 [>.............................] - ETA: 0s - loss: 0.9197 - accuracy: 0.6641 6/17 [=========>....................] - ETA: 0s - loss: 0.8745 - accuracy: 0.6982 11/17 [==================>...........] - ETA: 0s - loss: 0.8673 - accuracy: 0.7055 16/17 [===========================>..] - ETA: 0s - loss: 0.8634 - accuracy: 0.7076 17/17 [==============================] - 0s 11ms/step - loss: 0.8619 - accuracy: 0.7089 Epoch 11/100 1/17 [>.............................] - ETA: 0s - loss: 1.0099 - accuracy: 0.7188 6/17 [=========>....................] - ETA: 0s - loss: 0.8879 - accuracy: 0.7244 11/17 [==================>...........] - ETA: 0s - loss: 0.8496 - accuracy: 0.7363 15/17 [=========================>....] - ETA: 0s - loss: 0.8283 - accuracy: 0.7417 17/17 [==============================] - 0s 12ms/step - loss: 0.8184 - accuracy: 0.7439 Epoch 12/100 1/17 [>.............................] - ETA: 0s - loss: 0.8259 - accuracy: 0.7109 6/17 [=========>....................] - ETA: 0s - loss: 0.7784 - accuracy: 0.7414 11/17 [==================>...........] - ETA: 0s - loss: 0.7591 - accuracy: 0.7483 16/17 [===========================>..] - ETA: 0s - loss: 0.7423 - accuracy: 0.7551 17/17 [==============================] - 0s 12ms/step - loss: 0.7374 - accuracy: 0.7569 Epoch 13/100 1/17 [>.............................] - ETA: 0s - loss: 0.6955 - accuracy: 0.7500 6/17 [=========>....................] - ETA: 0s - loss: 0.6596 - accuracy: 0.7760 11/17 [==================>...........] - ETA: 0s - loss: 0.6494 - accuracy: 0.7854 16/17 [===========================>..] - ETA: 0s - loss: 0.6423 - accuracy: 0.7900 17/17 [==============================] - 0s 12ms/step - loss: 0.6403 - accuracy: 0.7911 Epoch 14/100 1/17 [>.............................] - ETA: 0s - loss: 0.6502 - accuracy: 0.7734 6/17 [=========>....................] - ETA: 0s - loss: 0.6497 - accuracy: 0.7858 11/17 [==================>...........] - ETA: 0s - loss: 0.6298 - accuracy: 0.7974 16/17 [===========================>..] - ETA: 0s - loss: 0.6209 - accuracy: 0.8018 17/17 [==============================] - 0s 12ms/step - loss: 0.6167 - accuracy: 0.8038 Epoch 15/100 1/17 [>.............................] - ETA: 0s - loss: 0.6496 - accuracy: 0.8125 6/17 [=========>....................] - ETA: 0s - loss: 0.5609 - accuracy: 0.8355 11/17 [==================>...........] - ETA: 0s - loss: 0.5419 - accuracy: 0.8397 16/17 [===========================>..] - ETA: 0s - loss: 0.5346 - accuracy: 0.8416 17/17 [==============================] - 0s 11ms/step - loss: 0.5324 - accuracy: 0.8423 Epoch 16/100 1/17 [>.............................] - ETA: 0s - loss: 0.5160 - accuracy: 0.8438 6/17 [=========>....................] - ETA: 0s - loss: 0.5105 - accuracy: 0.8399 11/17 [==================>...........] - ETA: 0s - loss: 0.4998 - accuracy: 0.8439 16/17 [===========================>..] - ETA: 0s - loss: 0.4931 - accuracy: 0.8450 17/17 [==============================] - 0s 11ms/step - loss: 0.4927 - accuracy: 0.8447 Epoch 17/100 1/17 [>.............................] - ETA: 0s - loss: 0.5245 - accuracy: 0.8203 6/17 [=========>....................] - ETA: 0s - loss: 0.4641 - accuracy: 0.8525 11/17 [==================>...........] - ETA: 0s - loss: 0.4567 - accuracy: 0.8571 15/17 [=========================>....] - ETA: 0s - loss: 0.4561 - accuracy: 0.8583 17/17 [==============================] - 0s 12ms/step - loss: 0.4550 - accuracy: 0.8589 Epoch 18/100 1/17 [>.............................] - ETA: 0s - loss: 0.4959 - accuracy: 0.8516 6/17 [=========>....................] - ETA: 0s - loss: 0.4513 - accuracy: 0.8627 11/17 [==================>...........] - ETA: 0s - loss: 0.4494 - accuracy: 0.8645 16/17 [===========================>..] - ETA: 0s - loss: 0.4433 - accuracy: 0.8676 17/17 [==============================] - 0s 12ms/step - loss: 0.4416 - accuracy: 0.8682 Epoch 19/100 1/17 [>.............................] - ETA: 0s - loss: 0.4211 - accuracy: 0.8672 6/17 [=========>....................] - ETA: 0s - loss: 0.3946 - accuracy: 0.8856 11/17 [==================>...........] - ETA: 0s - loss: 0.3856 - accuracy: 0.8875 16/17 [===========================>..] - ETA: 0s - loss: 0.3869 - accuracy: 0.8846 17/17 [==============================] - 0s 12ms/step - loss: 0.3883 - accuracy: 0.8835 Epoch 20/100 1/17 [>.............................] - ETA: 0s - loss: 0.3987 - accuracy: 0.8438 6/17 [=========>....................] - ETA: 0s - loss: 0.4040 - accuracy: 0.8551 11/17 [==================>...........] - ETA: 0s - loss: 0.3928 - accuracy: 0.8616 16/17 [===========================>..] - ETA: 0s - loss: 0.3865 - accuracy: 0.8666 17/17 [==============================] - 0s 11ms/step - loss: 0.3845 - accuracy: 0.8679 Epoch 21/100 1/17 [>.............................] - ETA: 0s - loss: 0.3630 - accuracy: 0.8828 6/17 [=========>....................] - ETA: 0s - loss: 0.3453 - accuracy: 0.8924 11/17 [==================>...........] - ETA: 0s - loss: 0.3466 - accuracy: 0.8929 16/17 [===========================>..] - ETA: 0s - loss: 0.3477 - accuracy: 0.8926 17/17 [==============================] - 0s 12ms/step - loss: 0.3481 - accuracy: 0.8922 Epoch 22/100 1/17 [>.............................] - ETA: 0s - loss: 0.3861 - accuracy: 0.8750 6/17 [=========>....................] - ETA: 0s - loss: 0.3267 - accuracy: 0.8976 10/17 [================>.............] - ETA: 0s - loss: 0.3257 - accuracy: 0.8985 15/17 [=========================>....] - ETA: 0s - loss: 0.3238 - accuracy: 0.8992 17/17 [==============================] - 0s 12ms/step - loss: 0.3229 - accuracy: 0.8996 Epoch 23/100 1/17 [>.............................] - ETA: 0s - loss: 0.2505 - accuracy: 0.9297 6/17 [=========>....................] - ETA: 0s - loss: 0.2773 - accuracy: 0.9223 11/17 [==================>...........] - ETA: 0s - loss: 0.2847 - accuracy: 0.9211 16/17 [===========================>..] - ETA: 0s - loss: 0.2869 - accuracy: 0.9201 17/17 [==============================] - 0s 11ms/step - loss: 0.2873 - accuracy: 0.9198 Epoch 24/100 1/17 [>.............................] - ETA: 0s - loss: 0.2271 - accuracy: 0.9297 6/17 [=========>....................] - ETA: 0s - loss: 0.2618 - accuracy: 0.9256 10/17 [================>.............] - ETA: 0s - loss: 0.2758 - accuracy: 0.9202 15/17 [=========================>....] - ETA: 0s - loss: 0.2835 - accuracy: 0.9164 17/17 [==============================] - 0s 12ms/step - loss: 0.2848 - accuracy: 0.9158 Epoch 25/100 1/17 [>.............................] - ETA: 0s - loss: 0.2458 - accuracy: 0.9062 6/17 [=========>....................] - ETA: 0s - loss: 0.2760 - accuracy: 0.9031 11/17 [==================>...........] - ETA: 0s - loss: 0.2806 - accuracy: 0.9059 16/17 [===========================>..] - ETA: 0s - loss: 0.2822 - accuracy: 0.9070 17/17 [==============================] - 0s 11ms/step - loss: 0.2820 - accuracy: 0.9076 Epoch 26/100 1/17 [>.............................] - ETA: 0s - loss: 0.2572 - accuracy: 0.9141 6/17 [=========>....................] - ETA: 0s - loss: 0.2824 - accuracy: 0.9143 11/17 [==================>...........] - ETA: 0s - loss: 0.2795 - accuracy: 0.9165 16/17 [===========================>..] - ETA: 0s - loss: 0.2756 - accuracy: 0.9183 17/17 [==============================] - 0s 11ms/step - loss: 0.2745 - accuracy: 0.9187 Epoch 27/100 1/17 [>.............................] - ETA: 0s - loss: 0.2871 - accuracy: 0.9141 6/17 [=========>....................] - ETA: 0s - loss: 0.2802 - accuracy: 0.9146 11/17 [==================>...........] - ETA: 0s - loss: 0.2690 - accuracy: 0.9203 16/17 [===========================>..] - ETA: 0s - loss: 0.2609 - accuracy: 0.9239 17/17 [==============================] - 0s 11ms/step - loss: 0.2591 - accuracy: 0.9247 Epoch 28/100 1/17 [>.............................] - ETA: 0s - loss: 0.2078 - accuracy: 0.9531 6/17 [=========>....................] - ETA: 0s - loss: 0.2247 - accuracy: 0.9448 11/17 [==================>...........] - ETA: 0s - loss: 0.2305 - accuracy: 0.9400 15/17 [=========================>....] - ETA: 0s - loss: 0.2302 - accuracy: 0.9387 17/17 [==============================] - 0s 12ms/step - loss: 0.2313 - accuracy: 0.9377 Epoch 29/100 1/17 [>.............................] - ETA: 0s - loss: 0.1805 - accuracy: 0.9609 6/17 [=========>....................] - ETA: 0s - loss: 0.2042 - accuracy: 0.9455 11/17 [==================>...........] - ETA: 0s - loss: 0.2147 - accuracy: 0.9412 16/17 [===========================>..] - ETA: 0s - loss: 0.2219 - accuracy: 0.9382 17/17 [==============================] - 0s 12ms/step - loss: 0.2242 - accuracy: 0.9371 Epoch 30/100 1/17 [>.............................] - ETA: 0s - loss: 0.1775 - accuracy: 0.9531 6/17 [=========>....................] - ETA: 0s - loss: 0.2107 - accuracy: 0.9393 11/17 [==================>...........] - ETA: 0s - loss: 0.2152 - accuracy: 0.9374 16/17 [===========================>..] - ETA: 0s - loss: 0.2165 - accuracy: 0.9364 17/17 [==============================] - 0s 12ms/step - loss: 0.2171 - accuracy: 0.9358 Epoch 31/100 1/17 [>.............................] - ETA: 0s - loss: 0.2359 - accuracy: 0.9297 6/17 [=========>....................] - ETA: 0s - loss: 0.2269 - accuracy: 0.9360 11/17 [==================>...........] - ETA: 0s - loss: 0.2304 - accuracy: 0.9363 16/17 [===========================>..] - ETA: 0s - loss: 0.2294 - accuracy: 0.9363 17/17 [==============================] - 0s 12ms/step - loss: 0.2285 - accuracy: 0.9365 Epoch 32/100 1/17 [>.............................] - ETA: 0s - loss: 0.2654 - accuracy: 0.8984 6/17 [=========>....................] - ETA: 0s - loss: 0.2375 - accuracy: 0.9173 11/17 [==================>...........] - ETA: 0s - loss: 0.2182 - accuracy: 0.9284 16/17 [===========================>..] - ETA: 0s - loss: 0.2096 - accuracy: 0.9348 17/17 [==============================] - 0s 11ms/step - loss: 0.2079 - accuracy: 0.9358 Epoch 33/100 1/17 [>.............................] - ETA: 0s - loss: 0.1747 - accuracy: 0.9453 6/17 [=========>....................] - ETA: 0s - loss: 0.1882 - accuracy: 0.9398 10/17 [================>.............] - ETA: 0s - loss: 0.1903 - accuracy: 0.9415 15/17 [=========================>....] - ETA: 0s - loss: 0.1883 - accuracy: 0.9427 17/17 [==============================] - 0s 12ms/step - loss: 0.1881 - accuracy: 0.9430 Epoch 34/100 1/17 [>.............................] - ETA: 0s - loss: 0.1808 - accuracy: 0.9375 6/17 [=========>....................] - ETA: 0s - loss: 0.1642 - accuracy: 0.9544 11/17 [==================>...........] - ETA: 0s - loss: 0.1638 - accuracy: 0.9578 16/17 [===========================>..] - ETA: 0s - loss: 0.1694 - accuracy: 0.9559 17/17 [==============================] - 0s 12ms/step - loss: 0.1703 - accuracy: 0.9556 Epoch 35/100 1/17 [>.............................] - ETA: 0s - loss: 0.1859 - accuracy: 0.9375 6/17 [=========>....................] - ETA: 0s - loss: 0.1785 - accuracy: 0.9383 11/17 [==================>...........] - ETA: 0s - loss: 0.1768 - accuracy: 0.9432 16/17 [===========================>..] - ETA: 0s - loss: 0.1755 - accuracy: 0.9458 17/17 [==============================] - 0s 11ms/step - loss: 0.1751 - accuracy: 0.9464 Epoch 36/100 1/17 [>.............................] - ETA: 0s - loss: 0.1922 - accuracy: 0.9297 6/17 [=========>....................] - ETA: 0s - loss: 0.1937 - accuracy: 0.9427 11/17 [==================>...........] - ETA: 0s - loss: 0.1890 - accuracy: 0.9472 16/17 [===========================>..] - ETA: 0s - loss: 0.1857 - accuracy: 0.9490 17/17 [==============================] - 0s 12ms/step - loss: 0.1843 - accuracy: 0.9495 Epoch 37/100 1/17 [>.............................] - ETA: 0s - loss: 0.1418 - accuracy: 0.9531 6/17 [=========>....................] - ETA: 0s - loss: 0.1484 - accuracy: 0.9620 11/17 [==================>...........] - ETA: 0s - loss: 0.1556 - accuracy: 0.9598 15/17 [=========================>....] - ETA: 0s - loss: 0.1574 - accuracy: 0.9590 17/17 [==============================] - 0s 12ms/step - loss: 0.1580 - accuracy: 0.9588 Epoch 38/100 1/17 [>.............................] - ETA: 0s - loss: 0.1935 - accuracy: 0.9453 6/17 [=========>....................] - ETA: 0s - loss: 0.1601 - accuracy: 0.9541 11/17 [==================>...........] - ETA: 0s - loss: 0.1582 - accuracy: 0.9548 16/17 [===========================>..] - ETA: 0s - loss: 0.1561 - accuracy: 0.9548 17/17 [==============================] - 0s 12ms/step - loss: 0.1557 - accuracy: 0.9548 Epoch 39/100 1/17 [>.............................] - ETA: 0s - loss: 0.1982 - accuracy: 0.9531 6/17 [=========>....................] - ETA: 0s - loss: 0.1597 - accuracy: 0.9604 10/17 [================>.............] - ETA: 0s - loss: 0.1625 - accuracy: 0.9569 15/17 [=========================>....] - ETA: 0s - loss: 0.1638 - accuracy: 0.9557 17/17 [==============================] - 0s 11ms/step - loss: 0.1647 - accuracy: 0.9548 Epoch 40/100 1/17 [>.............................] - ETA: 0s - loss: 0.1461 - accuracy: 0.9531 6/17 [=========>....................] - ETA: 0s - loss: 0.1590 - accuracy: 0.9505 11/17 [==================>...........] - ETA: 0s - loss: 0.1552 - accuracy: 0.9550 16/17 [===========================>..] - ETA: 0s - loss: 0.1504 - accuracy: 0.9579 17/17 [==============================] - 0s 11ms/step - loss: 0.1494 - accuracy: 0.9584 Epoch 41/100 1/17 [>.............................] - ETA: 0s - loss: 0.0598 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.1182 - accuracy: 0.9705 11/17 [==================>...........] - ETA: 0s - loss: 0.1275 - accuracy: 0.9679 16/17 [===========================>..] - ETA: 0s - loss: 0.1297 - accuracy: 0.9669 17/17 [==============================] - 0s 12ms/step - loss: 0.1299 - accuracy: 0.9665 Epoch 42/100 1/17 [>.............................] - ETA: 0s - loss: 0.1715 - accuracy: 0.9609 6/17 [=========>....................] - ETA: 0s - loss: 0.1422 - accuracy: 0.9693 11/17 [==================>...........] - ETA: 0s - loss: 0.1443 - accuracy: 0.9671 16/17 [===========================>..] - ETA: 0s - loss: 0.1435 - accuracy: 0.9660 17/17 [==============================] - 0s 12ms/step - loss: 0.1432 - accuracy: 0.9657 Epoch 43/100 1/17 [>.............................] - ETA: 0s - loss: 0.1445 - accuracy: 0.9688 6/17 [=========>....................] - ETA: 0s - loss: 0.1302 - accuracy: 0.9624 11/17 [==================>...........] - ETA: 0s - loss: 0.1299 - accuracy: 0.9610 15/17 [=========================>....] - ETA: 0s - loss: 0.1290 - accuracy: 0.9612 17/17 [==============================] - 0s 12ms/step - loss: 0.1293 - accuracy: 0.9613 Epoch 44/100 1/17 [>.............................] - ETA: 0s - loss: 0.0838 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0925 - accuracy: 0.9815 10/17 [================>.............] - ETA: 0s - loss: 0.0977 - accuracy: 0.9789 15/17 [=========================>....] - ETA: 0s - loss: 0.1028 - accuracy: 0.9769 17/17 [==============================] - 0s 12ms/step - loss: 0.1050 - accuracy: 0.9759 Epoch 45/100 1/17 [>.............................] - ETA: 0s - loss: 0.0793 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.1203 - accuracy: 0.9573 11/17 [==================>...........] - ETA: 0s - loss: 0.1269 - accuracy: 0.9567 16/17 [===========================>..] - ETA: 0s - loss: 0.1287 - accuracy: 0.9568 17/17 [==============================] - 0s 11ms/step - loss: 0.1292 - accuracy: 0.9569 Epoch 46/100 1/17 [>.............................] - ETA: 0s - loss: 0.0995 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.1154 - accuracy: 0.9715 11/17 [==================>...........] - ETA: 0s - loss: 0.1164 - accuracy: 0.9697 15/17 [=========================>....] - ETA: 0s - loss: 0.1172 - accuracy: 0.9681 17/17 [==============================] - 0s 11ms/step - loss: 0.1182 - accuracy: 0.9670 Epoch 47/100 1/17 [>.............................] - ETA: 0s - loss: 0.0841 - accuracy: 1.0000 6/17 [=========>....................] - ETA: 0s - loss: 0.1146 - accuracy: 0.9679 11/17 [==================>...........] - ETA: 0s - loss: 0.1185 - accuracy: 0.9645 15/17 [=========================>....] - ETA: 0s - loss: 0.1211 - accuracy: 0.9630 17/17 [==============================] - 0s 12ms/step - loss: 0.1220 - accuracy: 0.9626 Epoch 48/100 1/17 [>.............................] - ETA: 0s - loss: 0.1113 - accuracy: 0.9609 6/17 [=========>....................] - ETA: 0s - loss: 0.1320 - accuracy: 0.9509 11/17 [==================>...........] - ETA: 0s - loss: 0.1242 - accuracy: 0.9567 16/17 [===========================>..] - ETA: 0s - loss: 0.1216 - accuracy: 0.9592 17/17 [==============================] - 0s 11ms/step - loss: 0.1210 - accuracy: 0.9598 Epoch 49/100 1/17 [>.............................] - ETA: 0s - loss: 0.1114 - accuracy: 0.9609 6/17 [=========>....................] - ETA: 0s - loss: 0.1029 - accuracy: 0.9709 11/17 [==================>...........] - ETA: 0s - loss: 0.1022 - accuracy: 0.9724 16/17 [===========================>..] - ETA: 0s - loss: 0.1020 - accuracy: 0.9730 17/17 [==============================] - 0s 11ms/step - loss: 0.1015 - accuracy: 0.9733 Epoch 50/100 1/17 [>.............................] - ETA: 0s - loss: 0.0747 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.0984 - accuracy: 0.9751 11/17 [==================>...........] - ETA: 0s - loss: 0.1027 - accuracy: 0.9730 16/17 [===========================>..] - ETA: 0s - loss: 0.1039 - accuracy: 0.9717 17/17 [==============================] - 0s 12ms/step - loss: 0.1042 - accuracy: 0.9714 Epoch 51/100 1/17 [>.............................] - ETA: 0s - loss: 0.0892 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.1093 - accuracy: 0.9702 11/17 [==================>...........] - ETA: 0s - loss: 0.1094 - accuracy: 0.9706 16/17 [===========================>..] - ETA: 0s - loss: 0.1082 - accuracy: 0.9708 17/17 [==============================] - 0s 11ms/step - loss: 0.1079 - accuracy: 0.9707 Epoch 52/100 1/17 [>.............................] - ETA: 0s - loss: 0.1681 - accuracy: 0.9375 6/17 [=========>....................] - ETA: 0s - loss: 0.1280 - accuracy: 0.9589 11/17 [==================>...........] - ETA: 0s - loss: 0.1212 - accuracy: 0.9601 16/17 [===========================>..] - ETA: 0s - loss: 0.1184 - accuracy: 0.9611 17/17 [==============================] - 0s 11ms/step - loss: 0.1176 - accuracy: 0.9615 Epoch 53/100 1/17 [>.............................] - ETA: 0s - loss: 0.0574 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0886 - accuracy: 0.9743 11/17 [==================>...........] - ETA: 0s - loss: 0.0941 - accuracy: 0.9726 16/17 [===========================>..] - ETA: 0s - loss: 0.0951 - accuracy: 0.9727 17/17 [==============================] - 0s 11ms/step - loss: 0.0952 - accuracy: 0.9727 Epoch 54/100 1/17 [>.............................] - ETA: 0s - loss: 0.1014 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.1094 - accuracy: 0.9706 11/17 [==================>...........] - ETA: 0s - loss: 0.1091 - accuracy: 0.9692 16/17 [===========================>..] - ETA: 0s - loss: 0.1068 - accuracy: 0.9696 17/17 [==============================] - 0s 12ms/step - loss: 0.1062 - accuracy: 0.9697 Epoch 55/100 1/17 [>.............................] - ETA: 0s - loss: 0.0855 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0805 - accuracy: 0.9793 11/17 [==================>...........] - ETA: 0s - loss: 0.0855 - accuracy: 0.9762 16/17 [===========================>..] - ETA: 0s - loss: 0.0884 - accuracy: 0.9746 17/17 [==============================] - 0s 11ms/step - loss: 0.0891 - accuracy: 0.9743 Epoch 56/100 1/17 [>.............................] - ETA: 0s - loss: 0.0660 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.0845 - accuracy: 0.9787 11/17 [==================>...........] - ETA: 0s - loss: 0.0884 - accuracy: 0.9766 16/17 [===========================>..] - ETA: 0s - loss: 0.0901 - accuracy: 0.9766 17/17 [==============================] - 0s 11ms/step - loss: 0.0906 - accuracy: 0.9764 Epoch 57/100 1/17 [>.............................] - ETA: 0s - loss: 0.1431 - accuracy: 0.9375 6/17 [=========>....................] - ETA: 0s - loss: 0.0999 - accuracy: 0.9654 11/17 [==================>...........] - ETA: 0s - loss: 0.0947 - accuracy: 0.9687 16/17 [===========================>..] - ETA: 0s - loss: 0.0931 - accuracy: 0.9704 17/17 [==============================] - 0s 11ms/step - loss: 0.0931 - accuracy: 0.9707 Epoch 58/100 1/17 [>.............................] - ETA: 0s - loss: 0.0869 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0882 - accuracy: 0.9766 11/17 [==================>...........] - ETA: 0s - loss: 0.0894 - accuracy: 0.9762 16/17 [===========================>..] - ETA: 0s - loss: 0.0900 - accuracy: 0.9761 17/17 [==============================] - 0s 11ms/step - loss: 0.0901 - accuracy: 0.9762 Epoch 59/100 1/17 [>.............................] - ETA: 0s - loss: 0.0691 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0667 - accuracy: 0.9821 11/17 [==================>...........] - ETA: 0s - loss: 0.0707 - accuracy: 0.9805 16/17 [===========================>..] - ETA: 0s - loss: 0.0747 - accuracy: 0.9796 17/17 [==============================] - 0s 11ms/step - loss: 0.0759 - accuracy: 0.9794 Epoch 60/100 1/17 [>.............................] - ETA: 0s - loss: 0.0855 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0962 - accuracy: 0.9699 11/17 [==================>...........] - ETA: 0s - loss: 0.0970 - accuracy: 0.9690 16/17 [===========================>..] - ETA: 0s - loss: 0.0957 - accuracy: 0.9697 17/17 [==============================] - 0s 11ms/step - loss: 0.0954 - accuracy: 0.9700 Epoch 61/100 1/17 [>.............................] - ETA: 0s - loss: 0.0740 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0995 - accuracy: 0.9755 11/17 [==================>...........] - ETA: 0s - loss: 0.0949 - accuracy: 0.9768 15/17 [=========================>....] - ETA: 0s - loss: 0.0934 - accuracy: 0.9771 17/17 [==============================] - 0s 12ms/step - loss: 0.0933 - accuracy: 0.9769 Epoch 62/100 1/17 [>.............................] - ETA: 0s - loss: 0.1036 - accuracy: 0.9688 6/17 [=========>....................] - ETA: 0s - loss: 0.0821 - accuracy: 0.9765 11/17 [==================>...........] - ETA: 0s - loss: 0.0798 - accuracy: 0.9776 16/17 [===========================>..] - ETA: 0s - loss: 0.0795 - accuracy: 0.9782 17/17 [==============================] - 0s 11ms/step - loss: 0.0798 - accuracy: 0.9783 Epoch 63/100 1/17 [>.............................] - ETA: 0s - loss: 0.0554 - accuracy: 1.0000 6/17 [=========>....................] - ETA: 0s - loss: 0.0632 - accuracy: 0.9885 11/17 [==================>...........] - ETA: 0s - loss: 0.0664 - accuracy: 0.9861 16/17 [===========================>..] - ETA: 0s - loss: 0.0690 - accuracy: 0.9848 17/17 [==============================] - 0s 11ms/step - loss: 0.0695 - accuracy: 0.9845 Epoch 64/100 1/17 [>.............................] - ETA: 0s - loss: 0.0509 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0575 - accuracy: 0.9855 11/17 [==================>...........] - ETA: 0s - loss: 0.0633 - accuracy: 0.9835 16/17 [===========================>..] - ETA: 0s - loss: 0.0665 - accuracy: 0.9823 17/17 [==============================] - 0s 11ms/step - loss: 0.0670 - accuracy: 0.9822 Epoch 65/100 1/17 [>.............................] - ETA: 0s - loss: 0.0640 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.0758 - accuracy: 0.9808 11/17 [==================>...........] - ETA: 0s - loss: 0.0761 - accuracy: 0.9805 16/17 [===========================>..] - ETA: 0s - loss: 0.0754 - accuracy: 0.9806 17/17 [==============================] - 0s 11ms/step - loss: 0.0751 - accuracy: 0.9807 Epoch 66/100 1/17 [>.............................] - ETA: 0s - loss: 0.0475 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.0734 - accuracy: 0.9759 11/17 [==================>...........] - ETA: 0s - loss: 0.0754 - accuracy: 0.9766 16/17 [===========================>..] - ETA: 0s - loss: 0.0744 - accuracy: 0.9777 17/17 [==============================] - 0s 11ms/step - loss: 0.0743 - accuracy: 0.9781 Epoch 67/100 1/17 [>.............................] - ETA: 0s - loss: 0.0351 - accuracy: 1.0000 6/17 [=========>....................] - ETA: 0s - loss: 0.0537 - accuracy: 0.9878 11/17 [==================>...........] - ETA: 0s - loss: 0.0593 - accuracy: 0.9867 16/17 [===========================>..] - ETA: 0s - loss: 0.0613 - accuracy: 0.9858 17/17 [==============================] - 0s 11ms/step - loss: 0.0619 - accuracy: 0.9855 Epoch 68/100 1/17 [>.............................] - ETA: 0s - loss: 0.0721 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0682 - accuracy: 0.9837 11/17 [==================>...........] - ETA: 0s - loss: 0.0690 - accuracy: 0.9827 16/17 [===========================>..] - ETA: 0s - loss: 0.0684 - accuracy: 0.9822 17/17 [==============================] - 0s 12ms/step - loss: 0.0683 - accuracy: 0.9820 Epoch 69/100 1/17 [>.............................] - ETA: 0s - loss: 0.0600 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0647 - accuracy: 0.9828 11/17 [==================>...........] - ETA: 0s - loss: 0.0651 - accuracy: 0.9829 16/17 [===========================>..] - ETA: 0s - loss: 0.0664 - accuracy: 0.9824 17/17 [==============================] - 0s 11ms/step - loss: 0.0669 - accuracy: 0.9822 Epoch 70/100 1/17 [>.............................] - ETA: 0s - loss: 0.0889 - accuracy: 0.9688 6/17 [=========>....................] - ETA: 0s - loss: 0.0655 - accuracy: 0.9827 11/17 [==================>...........] - ETA: 0s - loss: 0.0640 - accuracy: 0.9838 16/17 [===========================>..] - ETA: 0s - loss: 0.0656 - accuracy: 0.9831 17/17 [==============================] - 0s 12ms/step - loss: 0.0658 - accuracy: 0.9830 Epoch 71/100 1/17 [>.............................] - ETA: 0s - loss: 0.0890 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0656 - accuracy: 0.9875 11/17 [==================>...........] - ETA: 0s - loss: 0.0659 - accuracy: 0.9858 16/17 [===========================>..] - ETA: 0s - loss: 0.0677 - accuracy: 0.9844 17/17 [==============================] - 0s 11ms/step - loss: 0.0681 - accuracy: 0.9841 Epoch 72/100 1/17 [>.............................] - ETA: 0s - loss: 0.1514 - accuracy: 0.9688 6/17 [=========>....................] - ETA: 0s - loss: 0.0995 - accuracy: 0.9734 11/17 [==================>...........] - ETA: 0s - loss: 0.0905 - accuracy: 0.9747 16/17 [===========================>..] - ETA: 0s - loss: 0.0866 - accuracy: 0.9757 17/17 [==============================] - 0s 11ms/step - loss: 0.0857 - accuracy: 0.9760 Epoch 73/100 1/17 [>.............................] - ETA: 0s - loss: 0.0751 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0782 - accuracy: 0.9721 11/17 [==================>...........] - ETA: 0s - loss: 0.0742 - accuracy: 0.9741 16/17 [===========================>..] - ETA: 0s - loss: 0.0711 - accuracy: 0.9761 17/17 [==============================] - 0s 12ms/step - loss: 0.0701 - accuracy: 0.9767 Epoch 74/100 1/17 [>.............................] - ETA: 0s - loss: 0.0384 - accuracy: 1.0000 6/17 [=========>....................] - ETA: 0s - loss: 0.0810 - accuracy: 0.9850 10/17 [================>.............] - ETA: 0s - loss: 0.0814 - accuracy: 0.9822 15/17 [=========================>....] - ETA: 0s - loss: 0.0820 - accuracy: 0.9803 17/17 [==============================] - 0s 12ms/step - loss: 0.0820 - accuracy: 0.9799 Epoch 75/100 1/17 [>.............................] - ETA: 0s - loss: 0.0682 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0573 - accuracy: 0.9868 11/17 [==================>...........] - ETA: 0s - loss: 0.0570 - accuracy: 0.9866 16/17 [===========================>..] - ETA: 0s - loss: 0.0604 - accuracy: 0.9856 17/17 [==============================] - 0s 11ms/step - loss: 0.0612 - accuracy: 0.9854 Epoch 76/100 1/17 [>.............................] - ETA: 0s - loss: 0.0712 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0669 - accuracy: 0.9796 11/17 [==================>...........] - ETA: 0s - loss: 0.0655 - accuracy: 0.9806 16/17 [===========================>..] - ETA: 0s - loss: 0.0661 - accuracy: 0.9806 17/17 [==============================] - 0s 11ms/step - loss: 0.0662 - accuracy: 0.9804 Epoch 77/100 1/17 [>.............................] - ETA: 0s - loss: 0.0804 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0687 - accuracy: 0.9790 10/17 [================>.............] - ETA: 0s - loss: 0.0686 - accuracy: 0.9797 15/17 [=========================>....] - ETA: 0s - loss: 0.0684 - accuracy: 0.9798 17/17 [==============================] - 0s 11ms/step - loss: 0.0678 - accuracy: 0.9800 Epoch 78/100 1/17 [>.............................] - ETA: 0s - loss: 0.0548 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.0597 - accuracy: 0.9857 11/17 [==================>...........] - ETA: 0s - loss: 0.0587 - accuracy: 0.9848 16/17 [===========================>..] - ETA: 0s - loss: 0.0589 - accuracy: 0.9840 17/17 [==============================] - 0s 11ms/step - loss: 0.0590 - accuracy: 0.9838 Epoch 79/100 1/17 [>.............................] - ETA: 0s - loss: 0.0645 - accuracy: 0.9766 7/17 [===========>..................] - ETA: 0s - loss: 0.0581 - accuracy: 0.9802 12/17 [====================>.........] - ETA: 0s - loss: 0.0617 - accuracy: 0.9805 17/17 [==============================] - ETA: 0s - loss: 0.0636 - accuracy: 0.9807 17/17 [==============================] - 0s 11ms/step - loss: 0.0640 - accuracy: 0.9807 Epoch 80/100 1/17 [>.............................] - ETA: 0s - loss: 0.0944 - accuracy: 0.9766 6/17 [=========>....................] - ETA: 0s - loss: 0.0637 - accuracy: 0.9848 11/17 [==================>...........] - ETA: 0s - loss: 0.0601 - accuracy: 0.9847 16/17 [===========================>..] - ETA: 0s - loss: 0.0598 - accuracy: 0.9846 17/17 [==============================] - 0s 11ms/step - loss: 0.0595 - accuracy: 0.9846 Epoch 81/100 1/17 [>.............................] - ETA: 0s - loss: 0.0520 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.0538 - accuracy: 0.9852 11/17 [==================>...........] - ETA: 0s - loss: 0.0553 - accuracy: 0.9827 16/17 [===========================>..] - ETA: 0s - loss: 0.0557 - accuracy: 0.9824 17/17 [==============================] - 0s 12ms/step - loss: 0.0556 - accuracy: 0.9824 Epoch 82/100 1/17 [>.............................] - ETA: 0s - loss: 0.0795 - accuracy: 0.9688 6/17 [=========>....................] - ETA: 0s - loss: 0.0647 - accuracy: 0.9810 11/17 [==================>...........] - ETA: 0s - loss: 0.0640 - accuracy: 0.9813 16/17 [===========================>..] - ETA: 0s - loss: 0.0632 - accuracy: 0.9815 17/17 [==============================] - 0s 11ms/step - loss: 0.0632 - accuracy: 0.9816 Epoch 83/100 1/17 [>.............................] - ETA: 0s - loss: 0.0582 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0601 - accuracy: 0.9852 11/17 [==================>...........] - ETA: 0s - loss: 0.0619 - accuracy: 0.9834 16/17 [===========================>..] - ETA: 0s - loss: 0.0632 - accuracy: 0.9825 17/17 [==============================] - 0s 11ms/step - loss: 0.0637 - accuracy: 0.9822 Epoch 84/100 1/17 [>.............................] - ETA: 0s - loss: 0.0336 - accuracy: 1.0000 6/17 [=========>....................] - ETA: 0s - loss: 0.0463 - accuracy: 0.9890 11/17 [==================>...........] - ETA: 0s - loss: 0.0506 - accuracy: 0.9863 16/17 [===========================>..] - ETA: 0s - loss: 0.0520 - accuracy: 0.9858 17/17 [==============================] - 0s 11ms/step - loss: 0.0524 - accuracy: 0.9858 Epoch 85/100 1/17 [>.............................] - ETA: 0s - loss: 0.1052 - accuracy: 0.9609 6/17 [=========>....................] - ETA: 0s - loss: 0.0732 - accuracy: 0.9740 11/17 [==================>...........] - ETA: 0s - loss: 0.0684 - accuracy: 0.9766 16/17 [===========================>..] - ETA: 0s - loss: 0.0671 - accuracy: 0.9777 17/17 [==============================] - 0s 11ms/step - loss: 0.0665 - accuracy: 0.9780 Epoch 86/100 1/17 [>.............................] - ETA: 0s - loss: 0.0576 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0555 - accuracy: 0.9821 11/17 [==================>...........] - ETA: 0s - loss: 0.0551 - accuracy: 0.9825 16/17 [===========================>..] - ETA: 0s - loss: 0.0578 - accuracy: 0.9809 17/17 [==============================] - 0s 11ms/step - loss: 0.0586 - accuracy: 0.9807 Epoch 87/100 1/17 [>.............................] - ETA: 0s - loss: 0.0879 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0624 - accuracy: 0.9834 11/17 [==================>...........] - ETA: 0s - loss: 0.0564 - accuracy: 0.9842 16/17 [===========================>..] - ETA: 0s - loss: 0.0546 - accuracy: 0.9844 17/17 [==============================] - 0s 11ms/step - loss: 0.0542 - accuracy: 0.9844 Epoch 88/100 1/17 [>.............................] - ETA: 0s - loss: 0.0474 - accuracy: 0.9844 7/17 [===========>..................] - ETA: 0s - loss: 0.0453 - accuracy: 0.9864 12/17 [====================>.........] - ETA: 0s - loss: 0.0485 - accuracy: 0.9851 17/17 [==============================] - ETA: 0s - loss: 0.0503 - accuracy: 0.9847 17/17 [==============================] - 0s 10ms/step - loss: 0.0506 - accuracy: 0.9847 Epoch 89/100 1/17 [>.............................] - ETA: 0s - loss: 0.0149 - accuracy: 1.0000 6/17 [=========>....................] - ETA: 0s - loss: 0.0334 - accuracy: 0.9918 11/17 [==================>...........] - ETA: 0s - loss: 0.0388 - accuracy: 0.9887 16/17 [===========================>..] - ETA: 0s - loss: 0.0434 - accuracy: 0.9870 17/17 [==============================] - 0s 11ms/step - loss: 0.0447 - accuracy: 0.9866 Epoch 90/100 1/17 [>.............................] - ETA: 0s - loss: 0.0478 - accuracy: 0.9766 7/17 [===========>..................] - ETA: 0s - loss: 0.0461 - accuracy: 0.9874 12/17 [====================>.........] - ETA: 0s - loss: 0.0459 - accuracy: 0.9882 17/17 [==============================] - ETA: 0s - loss: 0.0463 - accuracy: 0.9881 17/17 [==============================] - 0s 11ms/step - loss: 0.0465 - accuracy: 0.9880 Epoch 91/100 1/17 [>.............................] - ETA: 0s - loss: 0.0896 - accuracy: 0.9609 6/17 [=========>....................] - ETA: 0s - loss: 0.0741 - accuracy: 0.9717 11/17 [==================>...........] - ETA: 0s - loss: 0.0684 - accuracy: 0.9745 16/17 [===========================>..] - ETA: 0s - loss: 0.0658 - accuracy: 0.9752 17/17 [==============================] - 0s 11ms/step - loss: 0.0651 - accuracy: 0.9754 Epoch 92/100 1/17 [>.............................] - ETA: 0s - loss: 0.0638 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0684 - accuracy: 0.9794 11/17 [==================>...........] - ETA: 0s - loss: 0.0622 - accuracy: 0.9812 16/17 [===========================>..] - ETA: 0s - loss: 0.0590 - accuracy: 0.9823 17/17 [==============================] - 0s 11ms/step - loss: 0.0584 - accuracy: 0.9825 Epoch 93/100 1/17 [>.............................] - ETA: 0s - loss: 0.0721 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0522 - accuracy: 0.9897 11/17 [==================>...........] - ETA: 0s - loss: 0.0494 - accuracy: 0.9891 16/17 [===========================>..] - ETA: 0s - loss: 0.0482 - accuracy: 0.9891 17/17 [==============================] - 0s 11ms/step - loss: 0.0479 - accuracy: 0.9889 Epoch 94/100 1/17 [>.............................] - ETA: 0s - loss: 0.0442 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0570 - accuracy: 0.9824 11/17 [==================>...........] - ETA: 0s - loss: 0.0528 - accuracy: 0.9839 16/17 [===========================>..] - ETA: 0s - loss: 0.0519 - accuracy: 0.9843 17/17 [==============================] - 0s 11ms/step - loss: 0.0517 - accuracy: 0.9843 Epoch 95/100 1/17 [>.............................] - ETA: 0s - loss: 0.0523 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0603 - accuracy: 0.9828 11/17 [==================>...........] - ETA: 0s - loss: 0.0590 - accuracy: 0.9828 16/17 [===========================>..] - ETA: 0s - loss: 0.0570 - accuracy: 0.9835 17/17 [==============================] - 0s 11ms/step - loss: 0.0566 - accuracy: 0.9835 Epoch 96/100 1/17 [>.............................] - ETA: 0s - loss: 0.0564 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0515 - accuracy: 0.9809 11/17 [==================>...........] - ETA: 0s - loss: 0.0499 - accuracy: 0.9821 16/17 [===========================>..] - ETA: 0s - loss: 0.0480 - accuracy: 0.9840 17/17 [==============================] - 0s 11ms/step - loss: 0.0478 - accuracy: 0.9844 Epoch 97/100 1/17 [>.............................] - ETA: 0s - loss: 0.0644 - accuracy: 0.9688 6/17 [=========>....................] - ETA: 0s - loss: 0.0538 - accuracy: 0.9771 11/17 [==================>...........] - ETA: 0s - loss: 0.0526 - accuracy: 0.9789 16/17 [===========================>..] - ETA: 0s - loss: 0.0515 - accuracy: 0.9804 17/17 [==============================] - 0s 11ms/step - loss: 0.0511 - accuracy: 0.9809 Epoch 98/100 1/17 [>.............................] - ETA: 0s - loss: 0.0743 - accuracy: 0.9844 6/17 [=========>....................] - ETA: 0s - loss: 0.0550 - accuracy: 0.9866 11/17 [==================>...........] - ETA: 0s - loss: 0.0525 - accuracy: 0.9867 16/17 [===========================>..] - ETA: 0s - loss: 0.0515 - accuracy: 0.9868 17/17 [==============================] - 0s 11ms/step - loss: 0.0516 - accuracy: 0.9866 Epoch 99/100 1/17 [>.............................] - ETA: 0s - loss: 0.0333 - accuracy: 0.9922 6/17 [=========>....................] - ETA: 0s - loss: 0.0422 - accuracy: 0.9889 11/17 [==================>...........] - ETA: 0s - loss: 0.0450 - accuracy: 0.9889 15/17 [=========================>....] - ETA: 0s - loss: 0.0455 - accuracy: 0.9891 17/17 [==============================] - 0s 12ms/step - loss: 0.0454 - accuracy: 0.9891 Epoch 100/100 1/17 [>.............................] - ETA: 0s - loss: 0.0864 - accuracy: 0.9609 7/17 [===========>..................] - ETA: 0s - loss: 0.0586 - accuracy: 0.9777 12/17 [====================>.........] - ETA: 0s - loss: 0.0553 - accuracy: 0.9803 17/17 [==============================] - ETA: 0s - loss: 0.0533 - accuracy: 0.9820 17/17 [==============================] - 0s 11ms/step - loss: 0.0530 - accuracy: 0.9822 <tensorflow.python.keras.callbacks.History at 0x7fc82c426c88>
기본 MLP 모델 평가하기
# Helper function to print evaluation metrics.
def print_metrics(model_desc, eval_metrics):
"""Prints evaluation metrics.
Args:
model_desc: A description of the model.
eval_metrics: A dictionary mapping metric names to corresponding values. It
must contain the loss and accuracy metrics.
"""
print('\n')
print('Eval accuracy for ', model_desc, ': ', eval_metrics['accuracy'])
print('Eval loss for ', model_desc, ': ', eval_metrics['loss'])
if 'graph_loss' in eval_metrics:
print('Eval graph loss for ', model_desc, ': ', eval_metrics['graph_loss'])
eval_results = dict(
zip(base_model.metrics_names,
base_model.evaluate(test_dataset, steps=HPARAMS.eval_steps)))
print_metrics('Base MLP model', eval_results)
1/Unknown - 0s 155ms/step - loss: 1.3607 - accuracy: 0.7656 5/5 [==============================] - 0s 8ms/step - loss: 1.4110 - accuracy: 0.7866 Eval accuracy for Base MLP model : 0.7866184711456299 Eval loss for Base MLP model : 1.4110491275787354
그래프 정규화로 MLP 모델 훈련하기
그래프 정규화를 기존 tf.Keras.Model의 손실 항에 통합하려면 몇 줄의 코드만 있으면 됩니다. 기본 모델은 래핑되어 새로운 tf.Keras 서브 클래스 모델을 생성하며, 손실에는 그래프 정규화가 포함됩니다.
그래프 정규화의 점진적 이점을 평가하기 위해 새 기본 모델 인스턴스를 생성합니다. 이는 base_model이 이미 몇 번의 반복 동안 훈련되었으며, 이 훈련된 모델을 재사용하여 그래프 정규화 모델을 만드는 것은 base_model에 대한 공정한 비교가 되지 않기 때문입니다.
# Build a new base MLP model.
base_reg_model_tag, base_reg_model = 'FUNCTIONAL', make_mlp_functional_model(
HPARAMS)
# Wrap the base MLP model with graph regularization.
graph_reg_config = nsl.configs.make_graph_reg_config(
max_neighbors=HPARAMS.num_neighbors,
multiplier=HPARAMS.graph_regularization_multiplier,
distance_type=HPARAMS.distance_type,
sum_over_axis=-1)
graph_reg_model = nsl.keras.GraphRegularization(base_reg_model,
graph_reg_config)
graph_reg_model.compile(
optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
graph_reg_model.fit(train_dataset, epochs=HPARAMS.train_epochs, verbose=1)
Epoch 1/100
/tmpfs/src/tf_docs_env/lib/python3.6/site-packages/tensorflow/python/framework/indexed_slices.py:437: UserWarning: Converting sparse IndexedSlices(IndexedSlices(indices=Tensor("gradient_tape/GraphRegularization/graph_loss/Reshape_1:0", shape=(None,), dtype=int32), values=Tensor("gradient_tape/GraphRegularization/graph_loss/Reshape:0", shape=(None, 7), dtype=float32), dense_shape=Tensor("gradient_tape/GraphRegularization/graph_loss/Cast:0", shape=(2,), dtype=int32))) to a dense Tensor of unknown shape. This may consume a large amount of memory.
"shape. This may consume a large amount of memory." % value)
1/Unknown - 2s 2s/step - loss: 1.9376 - accuracy: 0.1406 - scaled_graph_loss: 0.0010
6/Unknown - 2s 11ms/step - loss: 1.9352 - accuracy: 0.1593 - scaled_graph_loss: 9.3120e-04
11/Unknown - 2s 11ms/step - loss: 1.9297 - accuracy: 0.1687 - scaled_graph_loss: 9.1192e-04
16/Unknown - 2s 11ms/step - loss: 1.9242 - accuracy: 0.1777 - scaled_graph_loss: 9.1595e-04
17/17 [==============================] - 2s 11ms/step - loss: 1.9225 - accuracy: 0.1801 - scaled_graph_loss: 9.2100e-04
Epoch 2/100
1/17 [>.............................] - ETA: 0s - loss: 1.8713 - accuracy: 0.2500 - scaled_graph_loss: 0.0013
6/17 [=========>....................] - ETA: 0s - loss: 1.8634 - accuracy: 0.2717 - scaled_graph_loss: 0.0012
11/17 [==================>...........] - ETA: 0s - loss: 1.8542 - accuracy: 0.2756 - scaled_graph_loss: 0.0013
16/17 [===========================>..] - ETA: 0s - loss: 1.8454 - accuracy: 0.2810 - scaled_graph_loss: 0.0014
17/17 [==============================] - 0s 11ms/step - loss: 1.8425 - accuracy: 0.2831 - scaled_graph_loss: 0.0014
Epoch 3/100
1/17 [>.............................] - ETA: 0s - loss: 1.7793 - accuracy: 0.2812 - scaled_graph_loss: 0.0020
6/17 [=========>....................] - ETA: 0s - loss: 1.7602 - accuracy: 0.3174 - scaled_graph_loss: 0.0022
11/17 [==================>...........] - ETA: 0s - loss: 1.7511 - accuracy: 0.3217 - scaled_graph_loss: 0.0023
16/17 [===========================>..] - ETA: 0s - loss: 1.7470 - accuracy: 0.3199 - scaled_graph_loss: 0.0025
17/17 [==============================] - 0s 11ms/step - loss: 1.7455 - accuracy: 0.3194 - scaled_graph_loss: 0.0025
Epoch 4/100
1/17 [>.............................] - ETA: 0s - loss: 1.6206 - accuracy: 0.4219 - scaled_graph_loss: 0.0038
6/17 [=========>....................] - ETA: 0s - loss: 1.6577 - accuracy: 0.3777 - scaled_graph_loss: 0.0039
11/17 [==================>...........] - ETA: 0s - loss: 1.6615 - accuracy: 0.3617 - scaled_graph_loss: 0.0040
16/17 [===========================>..] - ETA: 0s - loss: 1.6550 - accuracy: 0.3599 - scaled_graph_loss: 0.0041
17/17 [==============================] - 0s 11ms/step - loss: 1.6523 - accuracy: 0.3600 - scaled_graph_loss: 0.0042
Epoch 5/100
1/17 [>.............................] - ETA: 0s - loss: 1.6149 - accuracy: 0.3047 - scaled_graph_loss: 0.0057
6/17 [=========>....................] - ETA: 0s - loss: 1.5876 - accuracy: 0.3589 - scaled_graph_loss: 0.0057
11/17 [==================>...........] - ETA: 0s - loss: 1.5726 - accuracy: 0.3767 - scaled_graph_loss: 0.0059
16/17 [===========================>..] - ETA: 0s - loss: 1.5585 - accuracy: 0.3880 - scaled_graph_loss: 0.0060
17/17 [==============================] - 0s 11ms/step - loss: 1.5537 - accuracy: 0.3918 - scaled_graph_loss: 0.0061
Epoch 6/100
1/17 [>.............................] - ETA: 0s - loss: 1.4908 - accuracy: 0.4297 - scaled_graph_loss: 0.0081
6/17 [=========>....................] - ETA: 0s - loss: 1.4271 - accuracy: 0.4797 - scaled_graph_loss: 0.0084
11/17 [==================>...........] - ETA: 0s - loss: 1.4082 - accuracy: 0.4853 - scaled_graph_loss: 0.0087
16/17 [===========================>..] - ETA: 0s - loss: 1.3971 - accuracy: 0.4907 - scaled_graph_loss: 0.0090
17/17 [==============================] - 0s 12ms/step - loss: 1.3923 - accuracy: 0.4934 - scaled_graph_loss: 0.0091
Epoch 7/100
1/17 [>.............................] - ETA: 0s - loss: 1.3129 - accuracy: 0.5391 - scaled_graph_loss: 0.0122
6/17 [=========>....................] - ETA: 0s - loss: 1.2773 - accuracy: 0.5566 - scaled_graph_loss: 0.0123
11/17 [==================>...........] - ETA: 0s - loss: 1.2701 - accuracy: 0.5591 - scaled_graph_loss: 0.0127
16/17 [===========================>..] - ETA: 0s - loss: 1.2637 - accuracy: 0.5613 - scaled_graph_loss: 0.0129
17/17 [==============================] - 0s 11ms/step - loss: 1.2615 - accuracy: 0.5619 - scaled_graph_loss: 0.0131
Epoch 8/100
1/17 [>.............................] - ETA: 0s - loss: 1.1887 - accuracy: 0.6016 - scaled_graph_loss: 0.0145
6/17 [=========>....................] - ETA: 0s - loss: 1.1620 - accuracy: 0.6125 - scaled_graph_loss: 0.0163
11/17 [==================>...........] - ETA: 0s - loss: 1.1480 - accuracy: 0.6224 - scaled_graph_loss: 0.0164
16/17 [===========================>..] - ETA: 0s - loss: 1.1419 - accuracy: 0.6250 - scaled_graph_loss: 0.0166
17/17 [==============================] - 0s 12ms/step - loss: 1.1398 - accuracy: 0.6262 - scaled_graph_loss: 0.0167
Epoch 9/100
1/17 [>.............................] - ETA: 0s - loss: 1.0050 - accuracy: 0.7031 - scaled_graph_loss: 0.0232
6/17 [=========>....................] - ETA: 0s - loss: 1.0236 - accuracy: 0.6775 - scaled_graph_loss: 0.0211
11/17 [==================>...........] - ETA: 0s - loss: 1.0275 - accuracy: 0.6710 - scaled_graph_loss: 0.0210
16/17 [===========================>..] - ETA: 0s - loss: 1.0229 - accuracy: 0.6711 - scaled_graph_loss: 0.0211
17/17 [==============================] - 0s 11ms/step - loss: 1.0197 - accuracy: 0.6717 - scaled_graph_loss: 0.0212
Epoch 10/100
1/17 [>.............................] - ETA: 0s - loss: 0.8815 - accuracy: 0.7422 - scaled_graph_loss: 0.0244
6/17 [=========>....................] - ETA: 0s - loss: 0.9268 - accuracy: 0.6979 - scaled_graph_loss: 0.0246
11/17 [==================>...........] - ETA: 0s - loss: 0.9240 - accuracy: 0.6956 - scaled_graph_loss: 0.0246
16/17 [===========================>..] - ETA: 0s - loss: 0.9181 - accuracy: 0.6988 - scaled_graph_loss: 0.0246
17/17 [==============================] - 0s 11ms/step - loss: 0.9155 - accuracy: 0.7002 - scaled_graph_loss: 0.0247
Epoch 11/100
1/17 [>.............................] - ETA: 0s - loss: 0.7068 - accuracy: 0.7969 - scaled_graph_loss: 0.0261
6/17 [=========>....................] - ETA: 0s - loss: 0.7758 - accuracy: 0.7836 - scaled_graph_loss: 0.0258
11/17 [==================>...........] - ETA: 0s - loss: 0.7884 - accuracy: 0.7758 - scaled_graph_loss: 0.0259
16/17 [===========================>..] - ETA: 0s - loss: 0.7931 - accuracy: 0.7706 - scaled_graph_loss: 0.0258
17/17 [==============================] - 0s 11ms/step - loss: 0.7946 - accuracy: 0.7688 - scaled_graph_loss: 0.0258
Epoch 12/100
1/17 [>.............................] - ETA: 0s - loss: 0.7846 - accuracy: 0.8047 - scaled_graph_loss: 0.0252
6/17 [=========>....................] - ETA: 0s - loss: 0.7703 - accuracy: 0.7897 - scaled_graph_loss: 0.0260
11/17 [==================>...........] - ETA: 0s - loss: 0.7648 - accuracy: 0.7851 - scaled_graph_loss: 0.0266
16/17 [===========================>..] - ETA: 0s - loss: 0.7553 - accuracy: 0.7865 - scaled_graph_loss: 0.0271
17/17 [==============================] - 0s 11ms/step - loss: 0.7516 - accuracy: 0.7877 - scaled_graph_loss: 0.0272
Epoch 13/100
1/17 [>.............................] - ETA: 0s - loss: 0.6676 - accuracy: 0.8828 - scaled_graph_loss: 0.0217
6/17 [=========>....................] - ETA: 0s - loss: 0.6827 - accuracy: 0.8463 - scaled_graph_loss: 0.0255
11/17 [==================>...........] - ETA: 0s - loss: 0.6769 - accuracy: 0.8355 - scaled_graph_loss: 0.0268
16/17 [===========================>..] - ETA: 0s - loss: 0.6728 - accuracy: 0.8300 - scaled_graph_loss: 0.0274
17/17 [==============================] - 0s 12ms/step - loss: 0.6715 - accuracy: 0.8285 - scaled_graph_loss: 0.0276
Epoch 14/100
1/17 [>.............................] - ETA: 0s - loss: 0.5146 - accuracy: 0.8359 - scaled_graph_loss: 0.0326
6/17 [=========>....................] - ETA: 0s - loss: 0.5826 - accuracy: 0.8202 - scaled_graph_loss: 0.0315
11/17 [==================>...........] - ETA: 0s - loss: 0.6004 - accuracy: 0.8119 - scaled_graph_loss: 0.0315
15/17 [=========================>....] - ETA: 0s - loss: 0.6089 - accuracy: 0.8103 - scaled_graph_loss: 0.0317
17/17 [==============================] - 0s 11ms/step - loss: 0.6117 - accuracy: 0.8109 - scaled_graph_loss: 0.0317
Epoch 15/100
1/17 [>.............................] - ETA: 0s - loss: 0.6804 - accuracy: 0.8125 - scaled_graph_loss: 0.0310
6/17 [=========>....................] - ETA: 0s - loss: 0.6203 - accuracy: 0.8254 - scaled_graph_loss: 0.0313
11/17 [==================>...........] - ETA: 0s - loss: 0.5888 - accuracy: 0.8329 - scaled_graph_loss: 0.0313
16/17 [===========================>..] - ETA: 0s - loss: 0.5761 - accuracy: 0.8358 - scaled_graph_loss: 0.0313
17/17 [==============================] - 0s 11ms/step - loss: 0.5729 - accuracy: 0.8365 - scaled_graph_loss: 0.0313
Epoch 16/100
1/17 [>.............................] - ETA: 0s - loss: 0.5374 - accuracy: 0.8359 - scaled_graph_loss: 0.0336
6/17 [=========>....................] - ETA: 0s - loss: 0.5296 - accuracy: 0.8454 - scaled_graph_loss: 0.0331
11/17 [==================>...........] - ETA: 0s - loss: 0.5248 - accuracy: 0.8503 - scaled_graph_loss: 0.0328
16/17 [===========================>..] - ETA: 0s - loss: 0.5215 - accuracy: 0.8518 - scaled_graph_loss: 0.0327
17/17 [==============================] - 0s 12ms/step - loss: 0.5208 - accuracy: 0.8520 - scaled_graph_loss: 0.0327
Epoch 17/100
1/17 [>.............................] - ETA: 0s - loss: 0.3772 - accuracy: 0.9062 - scaled_graph_loss: 0.0297
6/17 [=========>....................] - ETA: 0s - loss: 0.4469 - accuracy: 0.8845 - scaled_graph_loss: 0.0297
11/17 [==================>...........] - ETA: 0s - loss: 0.4542 - accuracy: 0.8833 - scaled_graph_loss: 0.0300
16/17 [===========================>..] - ETA: 0s - loss: 0.4587 - accuracy: 0.8814 - scaled_graph_loss: 0.0305
17/17 [==============================] - 0s 12ms/step - loss: 0.4611 - accuracy: 0.8802 - scaled_graph_loss: 0.0307
Epoch 18/100
1/17 [>.............................] - ETA: 0s - loss: 0.4459 - accuracy: 0.9062 - scaled_graph_loss: 0.0365
6/17 [=========>....................] - ETA: 0s - loss: 0.4679 - accuracy: 0.8778 - scaled_graph_loss: 0.0333
11/17 [==================>...........] - ETA: 0s - loss: 0.4599 - accuracy: 0.8790 - scaled_graph_loss: 0.0325
16/17 [===========================>..] - ETA: 0s - loss: 0.4574 - accuracy: 0.8782 - scaled_graph_loss: 0.0324
17/17 [==============================] - 0s 11ms/step - loss: 0.4573 - accuracy: 0.8776 - scaled_graph_loss: 0.0324
Epoch 19/100
1/17 [>.............................] - ETA: 0s - loss: 0.4161 - accuracy: 0.9219 - scaled_graph_loss: 0.0330
6/17 [=========>....................] - ETA: 0s - loss: 0.3898 - accuracy: 0.9191 - scaled_graph_loss: 0.0327
11/17 [==================>...........] - ETA: 0s - loss: 0.3920 - accuracy: 0.9136 - scaled_graph_loss: 0.0323
16/17 [===========================>..] - ETA: 0s - loss: 0.3951 - accuracy: 0.9091 - scaled_graph_loss: 0.0320
17/17 [==============================] - 0s 11ms/step - loss: 0.3964 - accuracy: 0.9074 - scaled_graph_loss: 0.0320
Epoch 20/100
1/17 [>.............................] - ETA: 0s - loss: 0.3055 - accuracy: 0.9375 - scaled_graph_loss: 0.0310
6/17 [=========>....................] - ETA: 0s - loss: 0.3672 - accuracy: 0.9171 - scaled_graph_loss: 0.0327
11/17 [==================>...........] - ETA: 0s - loss: 0.3837 - accuracy: 0.9091 - scaled_graph_loss: 0.0338
16/17 [===========================>..] - ETA: 0s - loss: 0.3875 - accuracy: 0.9061 - scaled_graph_loss: 0.0338
17/17 [==============================] - 0s 11ms/step - loss: 0.3887 - accuracy: 0.9051 - scaled_graph_loss: 0.0337
Epoch 21/100
1/17 [>.............................] - ETA: 0s - loss: 0.3705 - accuracy: 0.9141 - scaled_graph_loss: 0.0344
6/17 [=========>....................] - ETA: 0s - loss: 0.3794 - accuracy: 0.8982 - scaled_graph_loss: 0.0367
11/17 [==================>...........] - ETA: 0s - loss: 0.3866 - accuracy: 0.8983 - scaled_graph_loss: 0.0360
16/17 [===========================>..] - ETA: 0s - loss: 0.3883 - accuracy: 0.8994 - scaled_graph_loss: 0.0352
17/17 [==============================] - 0s 11ms/step - loss: 0.3882 - accuracy: 0.8998 - scaled_graph_loss: 0.0350
Epoch 22/100
1/17 [>.............................] - ETA: 0s - loss: 0.3212 - accuracy: 0.9141 - scaled_graph_loss: 0.0364
5/17 [=======>......................] - ETA: 0s - loss: 0.3310 - accuracy: 0.9176 - scaled_graph_loss: 0.0356
10/17 [================>.............] - ETA: 0s - loss: 0.3426 - accuracy: 0.9121 - scaled_graph_loss: 0.0349
15/17 [=========================>....] - ETA: 0s - loss: 0.3454 - accuracy: 0.9094 - scaled_graph_loss: 0.0340
17/17 [==============================] - 0s 11ms/step - loss: 0.3457 - accuracy: 0.9086 - scaled_graph_loss: 0.0337
Epoch 23/100
1/17 [>.............................] - ETA: 0s - loss: 0.4233 - accuracy: 0.9062 - scaled_graph_loss: 0.0333
6/17 [=========>....................] - ETA: 0s - loss: 0.3976 - accuracy: 0.8968 - scaled_graph_loss: 0.0331
11/17 [==================>...........] - ETA: 0s - loss: 0.3773 - accuracy: 0.9006 - scaled_graph_loss: 0.0331
16/17 [===========================>..] - ETA: 0s - loss: 0.3692 - accuracy: 0.9014 - scaled_graph_loss: 0.0332
17/17 [==============================] - 0s 11ms/step - loss: 0.3666 - accuracy: 0.9020 - scaled_graph_loss: 0.0332
Epoch 24/100
1/17 [>.............................] - ETA: 0s - loss: 0.2855 - accuracy: 0.9453 - scaled_graph_loss: 0.0294
6/17 [=========>....................] - ETA: 0s - loss: 0.3148 - accuracy: 0.9211 - scaled_graph_loss: 0.0320
11/17 [==================>...........] - ETA: 0s - loss: 0.3255 - accuracy: 0.9169 - scaled_graph_loss: 0.0331
16/17 [===========================>..] - ETA: 0s - loss: 0.3258 - accuracy: 0.9173 - scaled_graph_loss: 0.0335
17/17 [==============================] - 0s 11ms/step - loss: 0.3249 - accuracy: 0.9178 - scaled_graph_loss: 0.0336
Epoch 25/100
1/17 [>.............................] - ETA: 0s - loss: 0.3132 - accuracy: 0.9219 - scaled_graph_loss: 0.0362
6/17 [=========>....................] - ETA: 0s - loss: 0.3215 - accuracy: 0.9072 - scaled_graph_loss: 0.0362
11/17 [==================>...........] - ETA: 0s - loss: 0.3120 - accuracy: 0.9102 - scaled_graph_loss: 0.0351
16/17 [===========================>..] - ETA: 0s - loss: 0.3086 - accuracy: 0.9122 - scaled_graph_loss: 0.0349
17/17 [==============================] - 0s 11ms/step - loss: 0.3070 - accuracy: 0.9131 - scaled_graph_loss: 0.0348
Epoch 26/100
1/17 [>.............................] - ETA: 0s - loss: 0.2629 - accuracy: 0.9297 - scaled_graph_loss: 0.0284
6/17 [=========>....................] - ETA: 0s - loss: 0.2644 - accuracy: 0.9327 - scaled_graph_loss: 0.0304
11/17 [==================>...........] - ETA: 0s - loss: 0.2685 - accuracy: 0.9336 - scaled_graph_loss: 0.0317
16/17 [===========================>..] - ETA: 0s - loss: 0.2696 - accuracy: 0.9342 - scaled_graph_loss: 0.0321
17/17 [==============================] - 0s 11ms/step - loss: 0.2703 - accuracy: 0.9342 - scaled_graph_loss: 0.0323
Epoch 27/100
1/17 [>.............................] - ETA: 0s - loss: 0.1986 - accuracy: 0.9844 - scaled_graph_loss: 0.0287
6/17 [=========>....................] - ETA: 0s - loss: 0.2637 - accuracy: 0.9443 - scaled_graph_loss: 0.0351
10/17 [================>.............] - ETA: 0s - loss: 0.2689 - accuracy: 0.9402 - scaled_graph_loss: 0.0351
15/17 [=========================>....] - ETA: 0s - loss: 0.2737 - accuracy: 0.9373 - scaled_graph_loss: 0.0347
17/17 [==============================] - 0s 11ms/step - loss: 0.2743 - accuracy: 0.9369 - scaled_graph_loss: 0.0346
Epoch 28/100
1/17 [>.............................] - ETA: 0s - loss: 0.2743 - accuracy: 0.9219 - scaled_graph_loss: 0.0310
6/17 [=========>....................] - ETA: 0s - loss: 0.2452 - accuracy: 0.9339 - scaled_graph_loss: 0.0318
11/17 [==================>...........] - ETA: 0s - loss: 0.2521 - accuracy: 0.9327 - scaled_graph_loss: 0.0329
16/17 [===========================>..] - ETA: 0s - loss: 0.2589 - accuracy: 0.9316 - scaled_graph_loss: 0.0334
17/17 [==============================] - 0s 11ms/step - loss: 0.2609 - accuracy: 0.9313 - scaled_graph_loss: 0.0334
Epoch 29/100
1/17 [>.............................] - ETA: 0s - loss: 0.2359 - accuracy: 0.9531 - scaled_graph_loss: 0.0367
6/17 [=========>....................] - ETA: 0s - loss: 0.2561 - accuracy: 0.9312 - scaled_graph_loss: 0.0341
11/17 [==================>...........] - ETA: 0s - loss: 0.2586 - accuracy: 0.9320 - scaled_graph_loss: 0.0337
16/17 [===========================>..] - ETA: 0s - loss: 0.2570 - accuracy: 0.9346 - scaled_graph_loss: 0.0335
17/17 [==============================] - 0s 12ms/step - loss: 0.2561 - accuracy: 0.9355 - scaled_graph_loss: 0.0334
Epoch 30/100
1/17 [>.............................] - ETA: 0s - loss: 0.2022 - accuracy: 0.9688 - scaled_graph_loss: 0.0306
6/17 [=========>....................] - ETA: 0s - loss: 0.2207 - accuracy: 0.9535 - scaled_graph_loss: 0.0308
11/17 [==================>...........] - ETA: 0s - loss: 0.2196 - accuracy: 0.9514 - scaled_graph_loss: 0.0311
15/17 [=========================>....] - ETA: 0s - loss: 0.2207 - accuracy: 0.9498 - scaled_graph_loss: 0.0315
17/17 [==============================] - 0s 11ms/step - loss: 0.2222 - accuracy: 0.9489 - scaled_graph_loss: 0.0318
Epoch 31/100
1/17 [>.............................] - ETA: 0s - loss: 0.1653 - accuracy: 0.9609 - scaled_graph_loss: 0.0292
6/17 [=========>....................] - ETA: 0s - loss: 0.1894 - accuracy: 0.9555 - scaled_graph_loss: 0.0309
11/17 [==================>...........] - ETA: 0s - loss: 0.1969 - accuracy: 0.9556 - scaled_graph_loss: 0.0314
16/17 [===========================>..] - ETA: 0s - loss: 0.2015 - accuracy: 0.9549 - scaled_graph_loss: 0.0322
17/17 [==============================] - 0s 11ms/step - loss: 0.2030 - accuracy: 0.9545 - scaled_graph_loss: 0.0324
Epoch 32/100
1/17 [>.............................] - ETA: 0s - loss: 0.2314 - accuracy: 0.9375 - scaled_graph_loss: 0.0298
6/17 [=========>....................] - ETA: 0s - loss: 0.2389 - accuracy: 0.9380 - scaled_graph_loss: 0.0321
11/17 [==================>...........] - ETA: 0s - loss: 0.2330 - accuracy: 0.9407 - scaled_graph_loss: 0.0326
16/17 [===========================>..] - ETA: 0s - loss: 0.2280 - accuracy: 0.9432 - scaled_graph_loss: 0.0329
17/17 [==============================] - 0s 11ms/step - loss: 0.2269 - accuracy: 0.9437 - scaled_graph_loss: 0.0329
Epoch 33/100
1/17 [>.............................] - ETA: 0s - loss: 0.2279 - accuracy: 0.9453 - scaled_graph_loss: 0.0422
6/17 [=========>....................] - ETA: 0s - loss: 0.2125 - accuracy: 0.9497 - scaled_graph_loss: 0.0366
11/17 [==================>...........] - ETA: 0s - loss: 0.2146 - accuracy: 0.9492 - scaled_graph_loss: 0.0355
16/17 [===========================>..] - ETA: 0s - loss: 0.2140 - accuracy: 0.9495 - scaled_graph_loss: 0.0352
17/17 [==============================] - 0s 11ms/step - loss: 0.2138 - accuracy: 0.9498 - scaled_graph_loss: 0.0351
Epoch 34/100
1/17 [>.............................] - ETA: 0s - loss: 0.1778 - accuracy: 0.9531 - scaled_graph_loss: 0.0373
6/17 [=========>....................] - ETA: 0s - loss: 0.2178 - accuracy: 0.9476 - scaled_graph_loss: 0.0352
11/17 [==================>...........] - ETA: 0s - loss: 0.2177 - accuracy: 0.9489 - scaled_graph_loss: 0.0349
16/17 [===========================>..] - ETA: 0s - loss: 0.2174 - accuracy: 0.9490 - scaled_graph_loss: 0.0348
17/17 [==============================] - 0s 11ms/step - loss: 0.2171 - accuracy: 0.9490 - scaled_graph_loss: 0.0347
Epoch 35/100
1/17 [>.............................] - ETA: 0s - loss: 0.2438 - accuracy: 0.9375 - scaled_graph_loss: 0.0338
6/17 [=========>....................] - ETA: 0s - loss: 0.2324 - accuracy: 0.9428 - scaled_graph_loss: 0.0349
11/17 [==================>...........] - ETA: 0s - loss: 0.2186 - accuracy: 0.9486 - scaled_graph_loss: 0.0344
16/17 [===========================>..] - ETA: 0s - loss: 0.2111 - accuracy: 0.9513 - scaled_graph_loss: 0.0343
17/17 [==============================] - 0s 11ms/step - loss: 0.2096 - accuracy: 0.9519 - scaled_graph_loss: 0.0344
Epoch 36/100
1/17 [>.............................] - ETA: 0s - loss: 0.1968 - accuracy: 0.9609 - scaled_graph_loss: 0.0297
6/17 [=========>....................] - ETA: 0s - loss: 0.2087 - accuracy: 0.9491 - scaled_graph_loss: 0.0336
11/17 [==================>...........] - ETA: 0s - loss: 0.2067 - accuracy: 0.9496 - scaled_graph_loss: 0.0349
16/17 [===========================>..] - ETA: 0s - loss: 0.2046 - accuracy: 0.9511 - scaled_graph_loss: 0.0351
17/17 [==============================] - 0s 11ms/step - loss: 0.2035 - accuracy: 0.9517 - scaled_graph_loss: 0.0350
Epoch 37/100
1/17 [>.............................] - ETA: 0s - loss: 0.2195 - accuracy: 0.9453 - scaled_graph_loss: 0.0363
6/17 [=========>....................] - ETA: 0s - loss: 0.1875 - accuracy: 0.9544 - scaled_graph_loss: 0.0330
11/17 [==================>...........] - ETA: 0s - loss: 0.1798 - accuracy: 0.9602 - scaled_graph_loss: 0.0326
16/17 [===========================>..] - ETA: 0s - loss: 0.1789 - accuracy: 0.9619 - scaled_graph_loss: 0.0328
17/17 [==============================] - 0s 11ms/step - loss: 0.1795 - accuracy: 0.9619 - scaled_graph_loss: 0.0330
Epoch 38/100
1/17 [>.............................] - ETA: 0s - loss: 0.1685 - accuracy: 0.9609 - scaled_graph_loss: 0.0400
6/17 [=========>....................] - ETA: 0s - loss: 0.1828 - accuracy: 0.9594 - scaled_graph_loss: 0.0360
11/17 [==================>...........] - ETA: 0s - loss: 0.1820 - accuracy: 0.9602 - scaled_graph_loss: 0.0351
16/17 [===========================>..] - ETA: 0s - loss: 0.1822 - accuracy: 0.9601 - scaled_graph_loss: 0.0347
17/17 [==============================] - 0s 11ms/step - loss: 0.1818 - accuracy: 0.9603 - scaled_graph_loss: 0.0346
Epoch 39/100
1/17 [>.............................] - ETA: 0s - loss: 0.1815 - accuracy: 0.9531 - scaled_graph_loss: 0.0322
6/17 [=========>....................] - ETA: 0s - loss: 0.1743 - accuracy: 0.9598 - scaled_graph_loss: 0.0325
10/17 [================>.............] - ETA: 0s - loss: 0.1749 - accuracy: 0.9606 - scaled_graph_loss: 0.0335
15/17 [=========================>....] - ETA: 0s - loss: 0.1762 - accuracy: 0.9603 - scaled_graph_loss: 0.0338
17/17 [==============================] - 0s 12ms/step - loss: 0.1778 - accuracy: 0.9596 - scaled_graph_loss: 0.0340
Epoch 40/100
1/17 [>.............................] - ETA: 0s - loss: 0.2141 - accuracy: 0.9531 - scaled_graph_loss: 0.0357
6/17 [=========>....................] - ETA: 0s - loss: 0.1774 - accuracy: 0.9622 - scaled_graph_loss: 0.0364
11/17 [==================>...........] - ETA: 0s - loss: 0.1778 - accuracy: 0.9604 - scaled_graph_loss: 0.0356
16/17 [===========================>..] - ETA: 0s - loss: 0.1788 - accuracy: 0.9596 - scaled_graph_loss: 0.0350
17/17 [==============================] - 0s 11ms/step - loss: 0.1788 - accuracy: 0.9597 - scaled_graph_loss: 0.0348
Epoch 41/100
1/17 [>.............................] - ETA: 0s - loss: 0.1363 - accuracy: 0.9922 - scaled_graph_loss: 0.0360
6/17 [=========>....................] - ETA: 0s - loss: 0.1452 - accuracy: 0.9782 - scaled_graph_loss: 0.0327
11/17 [==================>...........] - ETA: 0s - loss: 0.1547 - accuracy: 0.9735 - scaled_graph_loss: 0.0329
16/17 [===========================>..] - ETA: 0s - loss: 0.1595 - accuracy: 0.9705 - scaled_graph_loss: 0.0330
17/17 [==============================] - 0s 11ms/step - loss: 0.1604 - accuracy: 0.9699 - scaled_graph_loss: 0.0332
Epoch 42/100
1/17 [>.............................] - ETA: 0s - loss: 0.2170 - accuracy: 0.9453 - scaled_graph_loss: 0.0388
6/17 [=========>....................] - ETA: 0s - loss: 0.1811 - accuracy: 0.9610 - scaled_graph_loss: 0.0351
11/17 [==================>...........] - ETA: 0s - loss: 0.1731 - accuracy: 0.9658 - scaled_graph_loss: 0.0356
16/17 [===========================>..] - ETA: 0s - loss: 0.1699 - accuracy: 0.9678 - scaled_graph_loss: 0.0357
17/17 [==============================] - 0s 11ms/step - loss: 0.1692 - accuracy: 0.9682 - scaled_graph_loss: 0.0357
Epoch 43/100
1/17 [>.............................] - ETA: 0s - loss: 0.1367 - accuracy: 0.9922 - scaled_graph_loss: 0.0321
6/17 [=========>....................] - ETA: 0s - loss: 0.1510 - accuracy: 0.9785 - scaled_graph_loss: 0.0339
11/17 [==================>...........] - ETA: 0s - loss: 0.1502 - accuracy: 0.9767 - scaled_graph_loss: 0.0340
16/17 [===========================>..] - ETA: 0s - loss: 0.1502 - accuracy: 0.9750 - scaled_graph_loss: 0.0341
17/17 [==============================] - 0s 11ms/step - loss: 0.1506 - accuracy: 0.9744 - scaled_graph_loss: 0.0342
Epoch 44/100
1/17 [>.............................] - ETA: 0s - loss: 0.1790 - accuracy: 0.9766 - scaled_graph_loss: 0.0330
6/17 [=========>....................] - ETA: 0s - loss: 0.1885 - accuracy: 0.9609 - scaled_graph_loss: 0.0351
11/17 [==================>...........] - ETA: 0s - loss: 0.1820 - accuracy: 0.9617 - scaled_graph_loss: 0.0354
16/17 [===========================>..] - ETA: 0s - loss: 0.1772 - accuracy: 0.9627 - scaled_graph_loss: 0.0353
17/17 [==============================] - 0s 11ms/step - loss: 0.1763 - accuracy: 0.9628 - scaled_graph_loss: 0.0352
Epoch 45/100
1/17 [>.............................] - ETA: 0s - loss: 0.2210 - accuracy: 0.9688 - scaled_graph_loss: 0.0437
6/17 [=========>....................] - ETA: 0s - loss: 0.1930 - accuracy: 0.9647 - scaled_graph_loss: 0.0370
11/17 [==================>...........] - ETA: 0s - loss: 0.1788 - accuracy: 0.9661 - scaled_graph_loss: 0.0359
16/17 [===========================>..] - ETA: 0s - loss: 0.1739 - accuracy: 0.9658 - scaled_graph_loss: 0.0355
17/17 [==============================] - 0s 11ms/step - loss: 0.1721 - accuracy: 0.9657 - scaled_graph_loss: 0.0354
Epoch 46/100
1/17 [>.............................] - ETA: 0s - loss: 0.0928 - accuracy: 0.9844 - scaled_graph_loss: 0.0226
6/17 [=========>....................] - ETA: 0s - loss: 0.1277 - accuracy: 0.9755 - scaled_graph_loss: 0.0298
11/17 [==================>...........] - ETA: 0s - loss: 0.1347 - accuracy: 0.9737 - scaled_graph_loss: 0.0315
16/17 [===========================>..] - ETA: 0s - loss: 0.1377 - accuracy: 0.9729 - scaled_graph_loss: 0.0323
17/17 [==============================] - 0s 11ms/step - loss: 0.1386 - accuracy: 0.9726 - scaled_graph_loss: 0.0325
Epoch 47/100
1/17 [>.............................] - ETA: 0s - loss: 0.1282 - accuracy: 0.9688 - scaled_graph_loss: 0.0333
6/17 [=========>....................] - ETA: 0s - loss: 0.1414 - accuracy: 0.9648 - scaled_graph_loss: 0.0322
11/17 [==================>...........] - ETA: 0s - loss: 0.1473 - accuracy: 0.9643 - scaled_graph_loss: 0.0331
16/17 [===========================>..] - ETA: 0s - loss: 0.1461 - accuracy: 0.9664 - scaled_graph_loss: 0.0331
17/17 [==============================] - 0s 12ms/step - loss: 0.1458 - accuracy: 0.9669 - scaled_graph_loss: 0.0332
Epoch 48/100
1/17 [>.............................] - ETA: 0s - loss: 0.1176 - accuracy: 0.9766 - scaled_graph_loss: 0.0399
6/17 [=========>....................] - ETA: 0s - loss: 0.1104 - accuracy: 0.9869 - scaled_graph_loss: 0.0330
11/17 [==================>...........] - ETA: 0s - loss: 0.1158 - accuracy: 0.9845 - scaled_graph_loss: 0.0331
16/17 [===========================>..] - ETA: 0s - loss: 0.1199 - accuracy: 0.9821 - scaled_graph_loss: 0.0333
17/17 [==============================] - 0s 11ms/step - loss: 0.1211 - accuracy: 0.9815 - scaled_graph_loss: 0.0334
Epoch 49/100
1/17 [>.............................] - ETA: 0s - loss: 0.1235 - accuracy: 0.9766 - scaled_graph_loss: 0.0351
6/17 [=========>....................] - ETA: 0s - loss: 0.1307 - accuracy: 0.9754 - scaled_graph_loss: 0.0332
11/17 [==================>...........] - ETA: 0s - loss: 0.1272 - accuracy: 0.9782 - scaled_graph_loss: 0.0325
16/17 [===========================>..] - ETA: 0s - loss: 0.1275 - accuracy: 0.9787 - scaled_graph_loss: 0.0325
17/17 [==============================] - 0s 11ms/step - loss: 0.1281 - accuracy: 0.9786 - scaled_graph_loss: 0.0326
Epoch 50/100
1/17 [>.............................] - ETA: 0s - loss: 0.0950 - accuracy: 1.0000 - scaled_graph_loss: 0.0307
6/17 [=========>....................] - ETA: 0s - loss: 0.1147 - accuracy: 0.9899 - scaled_graph_loss: 0.0331
11/17 [==================>...........] - ETA: 0s - loss: 0.1241 - accuracy: 0.9848 - scaled_graph_loss: 0.0338
16/17 [===========================>..] - ETA: 0s - loss: 0.1277 - accuracy: 0.9822 - scaled_graph_loss: 0.0342
17/17 [==============================] - 0s 11ms/step - loss: 0.1285 - accuracy: 0.9814 - scaled_graph_loss: 0.0343
Epoch 51/100
1/17 [>.............................] - ETA: 0s - loss: 0.0952 - accuracy: 0.9844 - scaled_graph_loss: 0.0373
6/17 [=========>....................] - ETA: 0s - loss: 0.1222 - accuracy: 0.9770 - scaled_graph_loss: 0.0365
11/17 [==================>...........] - ETA: 0s - loss: 0.1258 - accuracy: 0.9768 - scaled_graph_loss: 0.0358
16/17 [===========================>..] - ETA: 0s - loss: 0.1304 - accuracy: 0.9753 - scaled_graph_loss: 0.0355
17/17 [==============================] - 0s 11ms/step - loss: 0.1317 - accuracy: 0.9748 - scaled_graph_loss: 0.0355
Epoch 52/100
1/17 [>.............................] - ETA: 0s - loss: 0.1488 - accuracy: 0.9688 - scaled_graph_loss: 0.0392
6/17 [=========>....................] - ETA: 0s - loss: 0.1431 - accuracy: 0.9714 - scaled_graph_loss: 0.0346
11/17 [==================>...........] - ETA: 0s - loss: 0.1431 - accuracy: 0.9707 - scaled_graph_loss: 0.0340
16/17 [===========================>..] - ETA: 0s - loss: 0.1425 - accuracy: 0.9705 - scaled_graph_loss: 0.0342
17/17 [==============================] - 0s 12ms/step - loss: 0.1420 - accuracy: 0.9706 - scaled_graph_loss: 0.0343
Epoch 53/100
1/17 [>.............................] - ETA: 0s - loss: 0.1613 - accuracy: 0.9688 - scaled_graph_loss: 0.0332
6/17 [=========>....................] - ETA: 0s - loss: 0.1376 - accuracy: 0.9751 - scaled_graph_loss: 0.0337
11/17 [==================>...........] - ETA: 0s - loss: 0.1391 - accuracy: 0.9721 - scaled_graph_loss: 0.0336
16/17 [===========================>..] - ETA: 0s - loss: 0.1396 - accuracy: 0.9715 - scaled_graph_loss: 0.0338
17/17 [==============================] - 0s 12ms/step - loss: 0.1395 - accuracy: 0.9715 - scaled_graph_loss: 0.0338
Epoch 54/100
1/17 [>.............................] - ETA: 0s - loss: 0.1148 - accuracy: 0.9922 - scaled_graph_loss: 0.0375
6/17 [=========>....................] - ETA: 0s - loss: 0.1235 - accuracy: 0.9808 - scaled_graph_loss: 0.0349
11/17 [==================>...........] - ETA: 0s - loss: 0.1253 - accuracy: 0.9772 - scaled_graph_loss: 0.0354
16/17 [===========================>..] - ETA: 0s - loss: 0.1258 - accuracy: 0.9761 - scaled_graph_loss: 0.0351
17/17 [==============================] - 0s 11ms/step - loss: 0.1260 - accuracy: 0.9758 - scaled_graph_loss: 0.0350
Epoch 55/100
1/17 [>.............................] - ETA: 0s - loss: 0.0901 - accuracy: 0.9922 - scaled_graph_loss: 0.0258
6/17 [=========>....................] - ETA: 0s - loss: 0.1198 - accuracy: 0.9780 - scaled_graph_loss: 0.0296
11/17 [==================>...........] - ETA: 0s - loss: 0.1243 - accuracy: 0.9776 - scaled_graph_loss: 0.0307
15/17 [=========================>....] - ETA: 0s - loss: 0.1259 - accuracy: 0.9775 - scaled_graph_loss: 0.0316
17/17 [==============================] - 0s 11ms/step - loss: 0.1261 - accuracy: 0.9778 - scaled_graph_loss: 0.0321
Epoch 56/100
1/17 [>.............................] - ETA: 0s - loss: 0.1181 - accuracy: 0.9844 - scaled_graph_loss: 0.0333
6/17 [=========>....................] - ETA: 0s - loss: 0.1175 - accuracy: 0.9827 - scaled_graph_loss: 0.0342
11/17 [==================>...........] - ETA: 0s - loss: 0.1197 - accuracy: 0.9808 - scaled_graph_loss: 0.0345
16/17 [===========================>..] - ETA: 0s - loss: 0.1203 - accuracy: 0.9796 - scaled_graph_loss: 0.0341
17/17 [==============================] - 0s 11ms/step - loss: 0.1205 - accuracy: 0.9793 - scaled_graph_loss: 0.0341
Epoch 57/100
1/17 [>.............................] - ETA: 0s - loss: 0.1004 - accuracy: 0.9922 - scaled_graph_loss: 0.0329
6/17 [=========>....................] - ETA: 0s - loss: 0.1079 - accuracy: 0.9832 - scaled_graph_loss: 0.0312
11/17 [==================>...........] - ETA: 0s - loss: 0.1138 - accuracy: 0.9818 - scaled_graph_loss: 0.0330
16/17 [===========================>..] - ETA: 0s - loss: 0.1164 - accuracy: 0.9815 - scaled_graph_loss: 0.0336
17/17 [==============================] - 0s 11ms/step - loss: 0.1170 - accuracy: 0.9814 - scaled_graph_loss: 0.0337
Epoch 58/100
1/17 [>.............................] - ETA: 0s - loss: 0.1054 - accuracy: 0.9844 - scaled_graph_loss: 0.0415
6/17 [=========>....................] - ETA: 0s - loss: 0.1196 - accuracy: 0.9708 - scaled_graph_loss: 0.0361
11/17 [==================>...........] - ETA: 0s - loss: 0.1212 - accuracy: 0.9706 - scaled_graph_loss: 0.0342
16/17 [===========================>..] - ETA: 0s - loss: 0.1220 - accuracy: 0.9712 - scaled_graph_loss: 0.0338
17/17 [==============================] - 0s 11ms/step - loss: 0.1223 - accuracy: 0.9715 - scaled_graph_loss: 0.0338
Epoch 59/100
1/17 [>.............................] - ETA: 0s - loss: 0.1173 - accuracy: 0.9688 - scaled_graph_loss: 0.0208
6/17 [=========>....................] - ETA: 0s - loss: 0.1207 - accuracy: 0.9713 - scaled_graph_loss: 0.0311
11/17 [==================>...........] - ETA: 0s - loss: 0.1204 - accuracy: 0.9721 - scaled_graph_loss: 0.0326
16/17 [===========================>..] - ETA: 0s - loss: 0.1185 - accuracy: 0.9734 - scaled_graph_loss: 0.0331
17/17 [==============================] - 0s 11ms/step - loss: 0.1181 - accuracy: 0.9737 - scaled_graph_loss: 0.0332
Epoch 60/100
1/17 [>.............................] - ETA: 0s - loss: 0.1053 - accuracy: 0.9766 - scaled_graph_loss: 0.0286
6/17 [=========>....................] - ETA: 0s - loss: 0.1044 - accuracy: 0.9840 - scaled_graph_loss: 0.0329
11/17 [==================>...........] - ETA: 0s - loss: 0.1081 - accuracy: 0.9832 - scaled_graph_loss: 0.0334
16/17 [===========================>..] - ETA: 0s - loss: 0.1101 - accuracy: 0.9828 - scaled_graph_loss: 0.0340
17/17 [==============================] - 0s 11ms/step - loss: 0.1104 - accuracy: 0.9827 - scaled_graph_loss: 0.0341
Epoch 61/100
1/17 [>.............................] - ETA: 0s - loss: 0.0593 - accuracy: 1.0000 - scaled_graph_loss: 0.0302
6/17 [=========>....................] - ETA: 0s - loss: 0.0810 - accuracy: 0.9908 - scaled_graph_loss: 0.0342
11/17 [==================>...........] - ETA: 0s - loss: 0.0864 - accuracy: 0.9870 - scaled_graph_loss: 0.0338
16/17 [===========================>..] - ETA: 0s - loss: 0.0905 - accuracy: 0.9847 - scaled_graph_loss: 0.0339
17/17 [==============================] - 0s 11ms/step - loss: 0.0919 - accuracy: 0.9840 - scaled_graph_loss: 0.0339
Epoch 62/100
1/17 [>.............................] - ETA: 0s - loss: 0.0795 - accuracy: 0.9844 - scaled_graph_loss: 0.0258
6/17 [=========>....................] - ETA: 0s - loss: 0.0992 - accuracy: 0.9788 - scaled_graph_loss: 0.0308
11/17 [==================>...........] - ETA: 0s - loss: 0.1001 - accuracy: 0.9818 - scaled_graph_loss: 0.0326
16/17 [===========================>..] - ETA: 0s - loss: 0.0999 - accuracy: 0.9835 - scaled_graph_loss: 0.0330
17/17 [==============================] - 0s 11ms/step - loss: 0.0999 - accuracy: 0.9838 - scaled_graph_loss: 0.0331
Epoch 63/100
1/17 [>.............................] - ETA: 0s - loss: 0.0849 - accuracy: 0.9844 - scaled_graph_loss: 0.0286
6/17 [=========>....................] - ETA: 0s - loss: 0.0960 - accuracy: 0.9860 - scaled_graph_loss: 0.0329
11/17 [==================>...........] - ETA: 0s - loss: 0.1044 - accuracy: 0.9844 - scaled_graph_loss: 0.0337
16/17 [===========================>..] - ETA: 0s - loss: 0.1072 - accuracy: 0.9835 - scaled_graph_loss: 0.0339
17/17 [==============================] - 0s 11ms/step - loss: 0.1078 - accuracy: 0.9833 - scaled_graph_loss: 0.0339
Epoch 64/100
1/17 [>.............................] - ETA: 0s - loss: 0.0896 - accuracy: 0.9844 - scaled_graph_loss: 0.0241
6/17 [=========>....................] - ETA: 0s - loss: 0.0899 - accuracy: 0.9877 - scaled_graph_loss: 0.0299
11/17 [==================>...........] - ETA: 0s - loss: 0.0968 - accuracy: 0.9859 - scaled_graph_loss: 0.0318
16/17 [===========================>..] - ETA: 0s - loss: 0.0991 - accuracy: 0.9855 - scaled_graph_loss: 0.0323
17/17 [==============================] - 0s 11ms/step - loss: 0.0994 - accuracy: 0.9854 - scaled_graph_loss: 0.0324
Epoch 65/100
1/17 [>.............................] - ETA: 0s - loss: 0.0885 - accuracy: 0.9922 - scaled_graph_loss: 0.0296
6/17 [=========>....................] - ETA: 0s - loss: 0.0995 - accuracy: 0.9829 - scaled_graph_loss: 0.0349
11/17 [==================>...........] - ETA: 0s - loss: 0.1013 - accuracy: 0.9823 - scaled_graph_loss: 0.0356
16/17 [===========================>..] - ETA: 0s - loss: 0.1017 - accuracy: 0.9820 - scaled_graph_loss: 0.0356
17/17 [==============================] - 0s 11ms/step - loss: 0.1016 - accuracy: 0.9820 - scaled_graph_loss: 0.0355
Epoch 66/100
1/17 [>.............................] - ETA: 0s - loss: 0.0967 - accuracy: 0.9844 - scaled_graph_loss: 0.0367
6/17 [=========>....................] - ETA: 0s - loss: 0.0957 - accuracy: 0.9870 - scaled_graph_loss: 0.0321
11/17 [==================>...........] - ETA: 0s - loss: 0.0960 - accuracy: 0.9863 - scaled_graph_loss: 0.0322
16/17 [===========================>..] - ETA: 0s - loss: 0.0961 - accuracy: 0.9861 - scaled_graph_loss: 0.0326
17/17 [==============================] - 0s 11ms/step - loss: 0.0962 - accuracy: 0.9859 - scaled_graph_loss: 0.0327
Epoch 67/100
1/17 [>.............................] - ETA: 0s - loss: 0.0854 - accuracy: 0.9922 - scaled_graph_loss: 0.0402
6/17 [=========>....................] - ETA: 0s - loss: 0.0963 - accuracy: 0.9889 - scaled_graph_loss: 0.0365
11/17 [==================>...........] - ETA: 0s - loss: 0.0990 - accuracy: 0.9856 - scaled_graph_loss: 0.0346
16/17 [===========================>..] - ETA: 0s - loss: 0.0997 - accuracy: 0.9848 - scaled_graph_loss: 0.0345
17/17 [==============================] - 0s 11ms/step - loss: 0.0999 - accuracy: 0.9846 - scaled_graph_loss: 0.0345
Epoch 68/100
1/17 [>.............................] - ETA: 0s - loss: 0.1096 - accuracy: 0.9844 - scaled_graph_loss: 0.0275
6/17 [=========>....................] - ETA: 0s - loss: 0.1051 - accuracy: 0.9829 - scaled_graph_loss: 0.0336
11/17 [==================>...........] - ETA: 0s - loss: 0.1030 - accuracy: 0.9831 - scaled_graph_loss: 0.0333
16/17 [===========================>..] - ETA: 0s - loss: 0.1032 - accuracy: 0.9824 - scaled_graph_loss: 0.0333
17/17 [==============================] - 0s 11ms/step - loss: 0.1032 - accuracy: 0.9823 - scaled_graph_loss: 0.0333
Epoch 69/100
1/17 [>.............................] - ETA: 0s - loss: 0.1295 - accuracy: 0.9766 - scaled_graph_loss: 0.0352
6/17 [=========>....................] - ETA: 0s - loss: 0.1036 - accuracy: 0.9838 - scaled_graph_loss: 0.0342
11/17 [==================>...........] - ETA: 0s - loss: 0.1047 - accuracy: 0.9830 - scaled_graph_loss: 0.0352
16/17 [===========================>..] - ETA: 0s - loss: 0.1038 - accuracy: 0.9827 - scaled_graph_loss: 0.0350
17/17 [==============================] - 0s 11ms/step - loss: 0.1035 - accuracy: 0.9828 - scaled_graph_loss: 0.0349
Epoch 70/100
1/17 [>.............................] - ETA: 0s - loss: 0.0942 - accuracy: 0.9922 - scaled_graph_loss: 0.0355
6/17 [=========>....................] - ETA: 0s - loss: 0.1045 - accuracy: 0.9850 - scaled_graph_loss: 0.0353
11/17 [==================>...........] - ETA: 0s - loss: 0.1053 - accuracy: 0.9832 - scaled_graph_loss: 0.0348
16/17 [===========================>..] - ETA: 0s - loss: 0.1055 - accuracy: 0.9828 - scaled_graph_loss: 0.0345
17/17 [==============================] - 0s 11ms/step - loss: 0.1052 - accuracy: 0.9828 - scaled_graph_loss: 0.0344
Epoch 71/100
1/17 [>.............................] - ETA: 0s - loss: 0.0961 - accuracy: 0.9922 - scaled_graph_loss: 0.0275
6/17 [=========>....................] - ETA: 0s - loss: 0.0960 - accuracy: 0.9870 - scaled_graph_loss: 0.0312
11/17 [==================>...........] - ETA: 0s - loss: 0.0938 - accuracy: 0.9873 - scaled_graph_loss: 0.0324
16/17 [===========================>..] - ETA: 0s - loss: 0.0925 - accuracy: 0.9873 - scaled_graph_loss: 0.0329
17/17 [==============================] - 0s 11ms/step - loss: 0.0921 - accuracy: 0.9874 - scaled_graph_loss: 0.0329
Epoch 72/100
1/17 [>.............................] - ETA: 0s - loss: 0.0762 - accuracy: 0.9922 - scaled_graph_loss: 0.0359
6/17 [=========>....................] - ETA: 0s - loss: 0.0814 - accuracy: 0.9913 - scaled_graph_loss: 0.0342
11/17 [==================>...........] - ETA: 0s - loss: 0.0862 - accuracy: 0.9886 - scaled_graph_loss: 0.0340
16/17 [===========================>..] - ETA: 0s - loss: 0.0899 - accuracy: 0.9870 - scaled_graph_loss: 0.0344
17/17 [==============================] - 0s 11ms/step - loss: 0.0907 - accuracy: 0.9867 - scaled_graph_loss: 0.0344
Epoch 73/100
1/17 [>.............................] - ETA: 0s - loss: 0.0951 - accuracy: 0.9922 - scaled_graph_loss: 0.0381
6/17 [=========>....................] - ETA: 0s - loss: 0.1009 - accuracy: 0.9842 - scaled_graph_loss: 0.0346
11/17 [==================>...........] - ETA: 0s - loss: 0.0991 - accuracy: 0.9843 - scaled_graph_loss: 0.0344
16/17 [===========================>..] - ETA: 0s - loss: 0.0984 - accuracy: 0.9849 - scaled_graph_loss: 0.0344
17/17 [==============================] - 0s 11ms/step - loss: 0.0982 - accuracy: 0.9851 - scaled_graph_loss: 0.0344
Epoch 74/100
1/17 [>.............................] - ETA: 0s - loss: 0.0709 - accuracy: 0.9844 - scaled_graph_loss: 0.0286
6/17 [=========>....................] - ETA: 0s - loss: 0.0795 - accuracy: 0.9844 - scaled_graph_loss: 0.0305
11/17 [==================>...........] - ETA: 0s - loss: 0.0859 - accuracy: 0.9827 - scaled_graph_loss: 0.0314
16/17 [===========================>..] - ETA: 0s - loss: 0.0887 - accuracy: 0.9818 - scaled_graph_loss: 0.0318
17/17 [==============================] - 0s 11ms/step - loss: 0.0892 - accuracy: 0.9817 - scaled_graph_loss: 0.0319
Epoch 75/100
1/17 [>.............................] - ETA: 0s - loss: 0.0960 - accuracy: 0.9922 - scaled_graph_loss: 0.0311
6/17 [=========>....................] - ETA: 0s - loss: 0.0959 - accuracy: 0.9875 - scaled_graph_loss: 0.0338
11/17 [==================>...........] - ETA: 0s - loss: 0.0974 - accuracy: 0.9841 - scaled_graph_loss: 0.0346
16/17 [===========================>..] - ETA: 0s - loss: 0.0966 - accuracy: 0.9837 - scaled_graph_loss: 0.0345
17/17 [==============================] - 0s 11ms/step - loss: 0.0966 - accuracy: 0.9835 - scaled_graph_loss: 0.0345
Epoch 76/100
1/17 [>.............................] - ETA: 0s - loss: 0.0602 - accuracy: 1.0000 - scaled_graph_loss: 0.0331
6/17 [=========>....................] - ETA: 0s - loss: 0.0786 - accuracy: 0.9923 - scaled_graph_loss: 0.0339
11/17 [==================>...........] - ETA: 0s - loss: 0.0842 - accuracy: 0.9902 - scaled_graph_loss: 0.0338
16/17 [===========================>..] - ETA: 0s - loss: 0.0879 - accuracy: 0.9891 - scaled_graph_loss: 0.0338
17/17 [==============================] - 0s 11ms/step - loss: 0.0888 - accuracy: 0.9888 - scaled_graph_loss: 0.0339
Epoch 77/100
1/17 [>.............................] - ETA: 0s - loss: 0.1211 - accuracy: 0.9766 - scaled_graph_loss: 0.0297
6/17 [=========>....................] - ETA: 0s - loss: 0.1032 - accuracy: 0.9780 - scaled_graph_loss: 0.0300
11/17 [==================>...........] - ETA: 0s - loss: 0.1027 - accuracy: 0.9788 - scaled_graph_loss: 0.0317
16/17 [===========================>..] - ETA: 0s - loss: 0.1024 - accuracy: 0.9796 - scaled_graph_loss: 0.0328
17/17 [==============================] - 0s 11ms/step - loss: 0.1019 - accuracy: 0.9800 - scaled_graph_loss: 0.0330
Epoch 78/100
1/17 [>.............................] - ETA: 0s - loss: 0.0729 - accuracy: 0.9844 - scaled_graph_loss: 0.0264
6/17 [=========>....................] - ETA: 0s - loss: 0.0913 - accuracy: 0.9782 - scaled_graph_loss: 0.0320
11/17 [==================>...........] - ETA: 0s - loss: 0.0930 - accuracy: 0.9791 - scaled_graph_loss: 0.0329
16/17 [===========================>..] - ETA: 0s - loss: 0.0933 - accuracy: 0.9802 - scaled_graph_loss: 0.0334
17/17 [==============================] - 0s 11ms/step - loss: 0.0931 - accuracy: 0.9806 - scaled_graph_loss: 0.0334
Epoch 79/100
1/17 [>.............................] - ETA: 0s - loss: 0.0906 - accuracy: 0.9844 - scaled_graph_loss: 0.0449
6/17 [=========>....................] - ETA: 0s - loss: 0.0888 - accuracy: 0.9836 - scaled_graph_loss: 0.0383
11/17 [==================>...........] - ETA: 0s - loss: 0.0878 - accuracy: 0.9843 - scaled_graph_loss: 0.0360
16/17 [===========================>..] - ETA: 0s - loss: 0.0865 - accuracy: 0.9847 - scaled_graph_loss: 0.0352
17/17 [==============================] - 0s 11ms/step - loss: 0.0862 - accuracy: 0.9848 - scaled_graph_loss: 0.0351
Epoch 80/100
1/17 [>.............................] - ETA: 0s - loss: 0.0865 - accuracy: 0.9922 - scaled_graph_loss: 0.0212
6/17 [=========>....................] - ETA: 0s - loss: 0.0800 - accuracy: 0.9897 - scaled_graph_loss: 0.0308
11/17 [==================>...........] - ETA: 0s - loss: 0.0839 - accuracy: 0.9875 - scaled_graph_loss: 0.0317
16/17 [===========================>..] - ETA: 0s - loss: 0.0846 - accuracy: 0.9872 - scaled_graph_loss: 0.0320
17/17 [==============================] - 0s 11ms/step - loss: 0.0852 - accuracy: 0.9870 - scaled_graph_loss: 0.0321
Epoch 81/100
1/17 [>.............................] - ETA: 0s - loss: 0.0965 - accuracy: 0.9922 - scaled_graph_loss: 0.0325
6/17 [=========>....................] - ETA: 0s - loss: 0.0940 - accuracy: 0.9876 - scaled_graph_loss: 0.0333
11/17 [==================>...........] - ETA: 0s - loss: 0.0929 - accuracy: 0.9849 - scaled_graph_loss: 0.0342
16/17 [===========================>..] - ETA: 0s - loss: 0.0928 - accuracy: 0.9840 - scaled_graph_loss: 0.0344
17/17 [==============================] - 0s 11ms/step - loss: 0.0927 - accuracy: 0.9837 - scaled_graph_loss: 0.0344
Epoch 82/100
1/17 [>.............................] - ETA: 0s - loss: 0.1080 - accuracy: 0.9766 - scaled_graph_loss: 0.0468
6/17 [=========>....................] - ETA: 0s - loss: 0.1017 - accuracy: 0.9788 - scaled_graph_loss: 0.0404
11/17 [==================>...........] - ETA: 0s - loss: 0.0991 - accuracy: 0.9808 - scaled_graph_loss: 0.0387
16/17 [===========================>..] - ETA: 0s - loss: 0.0961 - accuracy: 0.9822 - scaled_graph_loss: 0.0374
17/17 [==============================] - 0s 11ms/step - loss: 0.0954 - accuracy: 0.9826 - scaled_graph_loss: 0.0370
Epoch 83/100
1/17 [>.............................] - ETA: 0s - loss: 0.0816 - accuracy: 0.9922 - scaled_graph_loss: 0.0392
6/17 [=========>....................] - ETA: 0s - loss: 0.0822 - accuracy: 0.9897 - scaled_graph_loss: 0.0338
11/17 [==================>...........] - ETA: 0s - loss: 0.0821 - accuracy: 0.9900 - scaled_graph_loss: 0.0333
16/17 [===========================>..] - ETA: 0s - loss: 0.0811 - accuracy: 0.9904 - scaled_graph_loss: 0.0333
17/17 [==============================] - 0s 11ms/step - loss: 0.0810 - accuracy: 0.9904 - scaled_graph_loss: 0.0333
Epoch 84/100
1/17 [>.............................] - ETA: 0s - loss: 0.0945 - accuracy: 0.9844 - scaled_graph_loss: 0.0242
6/17 [=========>....................] - ETA: 0s - loss: 0.0782 - accuracy: 0.9880 - scaled_graph_loss: 0.0264
11/17 [==================>...........] - ETA: 0s - loss: 0.0802 - accuracy: 0.9877 - scaled_graph_loss: 0.0285
16/17 [===========================>..] - ETA: 0s - loss: 0.0823 - accuracy: 0.9874 - scaled_graph_loss: 0.0300
17/17 [==============================] - 0s 11ms/step - loss: 0.0827 - accuracy: 0.9874 - scaled_graph_loss: 0.0304
Epoch 85/100
1/17 [>.............................] - ETA: 0s - loss: 0.0836 - accuracy: 0.9922 - scaled_graph_loss: 0.0280
6/17 [=========>....................] - ETA: 0s - loss: 0.0927 - accuracy: 0.9861 - scaled_graph_loss: 0.0318
11/17 [==================>...........] - ETA: 0s - loss: 0.0914 - accuracy: 0.9852 - scaled_graph_loss: 0.0326
16/17 [===========================>..] - ETA: 0s - loss: 0.0903 - accuracy: 0.9853 - scaled_graph_loss: 0.0329
17/17 [==============================] - 0s 11ms/step - loss: 0.0898 - accuracy: 0.9854 - scaled_graph_loss: 0.0330
Epoch 86/100
1/17 [>.............................] - ETA: 0s - loss: 0.0460 - accuracy: 1.0000 - scaled_graph_loss: 0.0298
6/17 [=========>....................] - ETA: 0s - loss: 0.0707 - accuracy: 0.9898 - scaled_graph_loss: 0.0320
11/17 [==================>...........] - ETA: 0s - loss: 0.0793 - accuracy: 0.9870 - scaled_graph_loss: 0.0327
16/17 [===========================>..] - ETA: 0s - loss: 0.0823 - accuracy: 0.9865 - scaled_graph_loss: 0.0331
17/17 [==============================] - 0s 11ms/step - loss: 0.0829 - accuracy: 0.9864 - scaled_graph_loss: 0.0332
Epoch 87/100
1/17 [>.............................] - ETA: 0s - loss: 0.0676 - accuracy: 0.9922 - scaled_graph_loss: 0.0308
6/17 [=========>....................] - ETA: 0s - loss: 0.0769 - accuracy: 0.9902 - scaled_graph_loss: 0.0328
11/17 [==================>...........] - ETA: 0s - loss: 0.0783 - accuracy: 0.9897 - scaled_graph_loss: 0.0338
16/17 [===========================>..] - ETA: 0s - loss: 0.0784 - accuracy: 0.9893 - scaled_graph_loss: 0.0336
17/17 [==============================] - 0s 11ms/step - loss: 0.0784 - accuracy: 0.9893 - scaled_graph_loss: 0.0336
Epoch 88/100
1/17 [>.............................] - ETA: 0s - loss: 0.1109 - accuracy: 0.9688 - scaled_graph_loss: 0.0303
6/17 [=========>....................] - ETA: 0s - loss: 0.0875 - accuracy: 0.9849 - scaled_graph_loss: 0.0321
11/17 [==================>...........] - ETA: 0s - loss: 0.0834 - accuracy: 0.9868 - scaled_graph_loss: 0.0315
16/17 [===========================>..] - ETA: 0s - loss: 0.0815 - accuracy: 0.9875 - scaled_graph_loss: 0.0319
17/17 [==============================] - 0s 11ms/step - loss: 0.0811 - accuracy: 0.9876 - scaled_graph_loss: 0.0321
Epoch 89/100
1/17 [>.............................] - ETA: 0s - loss: 0.0753 - accuracy: 0.9922 - scaled_graph_loss: 0.0217
6/17 [=========>....................] - ETA: 0s - loss: 0.0795 - accuracy: 0.9907 - scaled_graph_loss: 0.0296
11/17 [==================>...........] - ETA: 0s - loss: 0.0809 - accuracy: 0.9893 - scaled_graph_loss: 0.0316
16/17 [===========================>..] - ETA: 0s - loss: 0.0812 - accuracy: 0.9888 - scaled_graph_loss: 0.0325
17/17 [==============================] - 0s 11ms/step - loss: 0.0811 - accuracy: 0.9887 - scaled_graph_loss: 0.0327
Epoch 90/100
1/17 [>.............................] - ETA: 0s - loss: 0.0914 - accuracy: 0.9844 - scaled_graph_loss: 0.0412
6/17 [=========>....................] - ETA: 0s - loss: 0.0856 - accuracy: 0.9848 - scaled_graph_loss: 0.0363
11/17 [==================>...........] - ETA: 0s - loss: 0.0825 - accuracy: 0.9854 - scaled_graph_loss: 0.0346
16/17 [===========================>..] - ETA: 0s - loss: 0.0815 - accuracy: 0.9856 - scaled_graph_loss: 0.0343
17/17 [==============================] - 0s 11ms/step - loss: 0.0813 - accuracy: 0.9856 - scaled_graph_loss: 0.0342
Epoch 91/100
1/17 [>.............................] - ETA: 0s - loss: 0.0812 - accuracy: 0.9922 - scaled_graph_loss: 0.0357
6/17 [=========>....................] - ETA: 0s - loss: 0.0689 - accuracy: 0.9929 - scaled_graph_loss: 0.0330
11/17 [==================>...........] - ETA: 0s - loss: 0.0709 - accuracy: 0.9912 - scaled_graph_loss: 0.0329
16/17 [===========================>..] - ETA: 0s - loss: 0.0750 - accuracy: 0.9900 - scaled_graph_loss: 0.0331
17/17 [==============================] - 0s 11ms/step - loss: 0.0765 - accuracy: 0.9896 - scaled_graph_loss: 0.0333
Epoch 92/100
1/17 [>.............................] - ETA: 0s - loss: 0.0862 - accuracy: 0.9766 - scaled_graph_loss: 0.0312
6/17 [=========>....................] - ETA: 0s - loss: 0.0872 - accuracy: 0.9798 - scaled_graph_loss: 0.0360
11/17 [==================>...........] - ETA: 0s - loss: 0.0880 - accuracy: 0.9811 - scaled_graph_loss: 0.0344
16/17 [===========================>..] - ETA: 0s - loss: 0.0887 - accuracy: 0.9812 - scaled_graph_loss: 0.0342
17/17 [==============================] - 0s 11ms/step - loss: 0.0888 - accuracy: 0.9814 - scaled_graph_loss: 0.0342
Epoch 93/100
1/17 [>.............................] - ETA: 0s - loss: 0.0543 - accuracy: 1.0000 - scaled_graph_loss: 0.0335
6/17 [=========>....................] - ETA: 0s - loss: 0.0659 - accuracy: 0.9885 - scaled_graph_loss: 0.0317
11/17 [==================>...........] - ETA: 0s - loss: 0.0733 - accuracy: 0.9858 - scaled_graph_loss: 0.0322
16/17 [===========================>..] - ETA: 0s - loss: 0.0784 - accuracy: 0.9845 - scaled_graph_loss: 0.0328
17/17 [==============================] - 0s 11ms/step - loss: 0.0796 - accuracy: 0.9843 - scaled_graph_loss: 0.0329
Epoch 94/100
1/17 [>.............................] - ETA: 0s - loss: 0.0891 - accuracy: 0.9844 - scaled_graph_loss: 0.0377
6/17 [=========>....................] - ETA: 0s - loss: 0.0795 - accuracy: 0.9837 - scaled_graph_loss: 0.0329
11/17 [==================>...........] - ETA: 0s - loss: 0.0780 - accuracy: 0.9852 - scaled_graph_loss: 0.0329
16/17 [===========================>..] - ETA: 0s - loss: 0.0784 - accuracy: 0.9857 - scaled_graph_loss: 0.0331
17/17 [==============================] - 0s 11ms/step - loss: 0.0784 - accuracy: 0.9859 - scaled_graph_loss: 0.0333
Epoch 95/100
1/17 [>.............................] - ETA: 0s - loss: 0.0768 - accuracy: 0.9922 - scaled_graph_loss: 0.0293
6/17 [=========>....................] - ETA: 0s - loss: 0.0810 - accuracy: 0.9905 - scaled_graph_loss: 0.0332
11/17 [==================>...........] - ETA: 0s - loss: 0.0831 - accuracy: 0.9888 - scaled_graph_loss: 0.0334
16/17 [===========================>..] - ETA: 0s - loss: 0.0839 - accuracy: 0.9883 - scaled_graph_loss: 0.0332
17/17 [==============================] - 0s 11ms/step - loss: 0.0842 - accuracy: 0.9882 - scaled_graph_loss: 0.0332
Epoch 96/100
1/17 [>.............................] - ETA: 0s - loss: 0.0622 - accuracy: 1.0000 - scaled_graph_loss: 0.0283
6/17 [=========>....................] - ETA: 0s - loss: 0.0642 - accuracy: 0.9975 - scaled_graph_loss: 0.0307
11/17 [==================>...........] - ETA: 0s - loss: 0.0712 - accuracy: 0.9936 - scaled_graph_loss: 0.0328
16/17 [===========================>..] - ETA: 0s - loss: 0.0732 - accuracy: 0.9924 - scaled_graph_loss: 0.0336
17/17 [==============================] - 0s 11ms/step - loss: 0.0739 - accuracy: 0.9920 - scaled_graph_loss: 0.0337
Epoch 97/100
1/17 [>.............................] - ETA: 0s - loss: 0.1011 - accuracy: 0.9688 - scaled_graph_loss: 0.0368
6/17 [=========>....................] - ETA: 0s - loss: 0.0876 - accuracy: 0.9803 - scaled_graph_loss: 0.0369
10/17 [================>.............] - ETA: 0s - loss: 0.0833 - accuracy: 0.9833 - scaled_graph_loss: 0.0358
15/17 [=========================>....] - ETA: 0s - loss: 0.0819 - accuracy: 0.9850 - scaled_graph_loss: 0.0351
17/17 [==============================] - 0s 12ms/step - loss: 0.0810 - accuracy: 0.9857 - scaled_graph_loss: 0.0347
Epoch 98/100
1/17 [>.............................] - ETA: 0s - loss: 0.0850 - accuracy: 0.9922 - scaled_graph_loss: 0.0397
6/17 [=========>....................] - ETA: 0s - loss: 0.0867 - accuracy: 0.9889 - scaled_graph_loss: 0.0369
11/17 [==================>...........] - ETA: 0s - loss: 0.0870 - accuracy: 0.9877 - scaled_graph_loss: 0.0361
16/17 [===========================>..] - ETA: 0s - loss: 0.0862 - accuracy: 0.9872 - scaled_graph_loss: 0.0357
17/17 [==============================] - 0s 11ms/step - loss: 0.0856 - accuracy: 0.9871 - scaled_graph_loss: 0.0356
Epoch 99/100
1/17 [>.............................] - ETA: 0s - loss: 0.0718 - accuracy: 0.9922 - scaled_graph_loss: 0.0202
6/17 [=========>....................] - ETA: 0s - loss: 0.0782 - accuracy: 0.9890 - scaled_graph_loss: 0.0275
11/17 [==================>...........] - ETA: 0s - loss: 0.0782 - accuracy: 0.9885 - scaled_graph_loss: 0.0287
16/17 [===========================>..] - ETA: 0s - loss: 0.0803 - accuracy: 0.9880 - scaled_graph_loss: 0.0301
17/17 [==============================] - 0s 11ms/step - loss: 0.0810 - accuracy: 0.9879 - scaled_graph_loss: 0.0305
Epoch 100/100
1/17 [>.............................] - ETA: 0s - loss: 0.0668 - accuracy: 0.9922 - scaled_graph_loss: 0.0353
6/17 [=========>....................] - ETA: 0s - loss: 0.0727 - accuracy: 0.9888 - scaled_graph_loss: 0.0348
11/17 [==================>...........] - ETA: 0s - loss: 0.0734 - accuracy: 0.9881 - scaled_graph_loss: 0.0341
16/17 [===========================>..] - ETA: 0s - loss: 0.0736 - accuracy: 0.9881 - scaled_graph_loss: 0.0339
17/17 [==============================] - 0s 11ms/step - loss: 0.0735 - accuracy: 0.9882 - scaled_graph_loss: 0.0339
<tensorflow.python.keras.callbacks.History at 0x7fc8206484a8>
그래프 정규화로 MLP 모델 평가하기
eval_results = dict(
zip(graph_reg_model.metrics_names,
graph_reg_model.evaluate(test_dataset, steps=HPARAMS.eval_steps)))
print_metrics('MLP + graph regularization', eval_results)
1/Unknown - 0s 200ms/step - loss: 1.3113 - accuracy: 0.7734 5/5 [==============================] - 0s 8ms/step - loss: 1.3513 - accuracy: 0.7946 Eval accuracy for MLP + graph regularization : 0.8010849952697754 Eval loss for MLP + graph regularization : 1.349246859550476
그래프 정규화 모델의 정확성은 기본 모델(base_model)보다 약 2~3% 높습니다.
결론
Neural Structured Learning(NSL) 프레임워크를 사용하여 자연 인용 그래프(Cora)에서 문서 분류를 위해 그래프 정규화를 사용하는 방법을 시연했습니다. 고급 튜토리얼에는 그래프 정규화로 신경망을 훈련하기 전에 샘플 임베딩을 기반으로 그래프를 합성하는 것이 포함됩니다. 이 접근 방식은 입력에 명시적 그래프가 포함되지 않은 경우 유용합니다.
사용자가 감독의 양을 변경하고 그래프 정규화를 위해 다양한 신경 아키텍처를 시도하여 추가 실험을 할 것을 권장합니다.