لماذا دقة نموذج Keras الخاص بي دوماً يساوي 0 وقيمة الخطأ كبيرة جداً عند التدريب

Meezo ML · 17 يوليو 2021

قمت ببناء نموذج في كيراس، لكن الدقة دوماً تساوي 0 وقيمة الخطأ كبيرة جداً، ما السبب؟

from keras.datasets import boston_housing
(train_data, train_targets), (test_data, test_targets) =boston_housing.load_data()
mean = train_data.mean(axis=0)
train_data -= mean
std = train_data.std(axis=0)
train_data /= std
test_data -= mean
test_data /= std
from keras import models
from keras import layers
model = models.Sequential()
model.add(layers.Dense(64, activation='relu',
input_shape=(train_data.shape[1],)))
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(1))
model.compile(optimizer='rmsprop', loss='mse', metrics=['acc'])
history = model.fit(train_data, train_targets,epochs=14, batch_size=64, verbose=1)
--------------------------------------------------------------------------------------
Epoch 1/14
7/7 [==============================] - 1s 4ms/step - loss: 540.0954 - acc: 0.0000e+00
Epoch 2/14
7/7 [==============================] - 0s 5ms/step - loss: 516.8516 - acc: 0.0000e+00
Epoch 3/14
7/7 [==============================] - 0s 4ms/step - loss: 463.7306 - acc: 0.0000e+00
Epoch 4/14
7/7 [==============================] - 0s 3ms/step - loss: 413.9145 - acc: 0.0000e+00
Epoch 5/14
7/7 [==============================] - 0s 2ms/step - loss: 374.7838 - acc: 0.0000e+00
Epoch 6/14
7/7 [==============================] - 0s 2ms/step - loss: 326.2928 - acc: 0.0000e+00
Epoch 7/14
7/7 [==============================] - 0s 3ms/step - loss: 284.4119 - acc: 0.0000e+00
Epoch 8/14
7/7 [==============================] - 0s 2ms/step - loss: 213.6348 - acc: 0.0000e+00
Epoch 9/14
7/7 [==============================] - 0s 3ms/step - loss: 161.0196 - acc: 0.0000e+00
Epoch 10/14
7/7 [==============================] - 0s 2ms/step - loss: 123.5639 - acc: 0.0000e+00
Epoch 11/14
7/7 [==============================] - 0s 3ms/step - loss: 102.8886 - acc: 0.0000e+00
Epoch 12/14
7/7 [==============================] - 0s 3ms/step - loss: 76.3049 - acc: 0.0000e+00
Epoch 13/14
7/7 [==============================] - 0s 3ms/step - loss: 65.1841 - acc: 0.0000e+00
Epoch 14/14
7/7 [==============================] - 0s 3ms/step - loss: 44.9363 - acc: 0.0000e+00

Ali Haidar Ahmad · 17 يوليو 2021

إن معيار قياس كفاءة النموذج الذي تستخدمه لنموذجك هو metrics=['accuracy'] وهو يتوافق مع مهام التصنيف، والمهمة التي لديك هي مهمة توقع لذا يجب عليك استخدام معيار يتناسب مع نوع المهمة مثل MSE أو MAE وبالتالي يجب يصبح الكود كالتالي:

from keras.datasets import boston_housing
(train_data, train_targets), (test_data, test_targets) =boston_housing.load_data()
mean = train_data.mean(axis=0)
train_data -= mean
std = train_data.std(axis=0)
train_data /= std
test_data -= mean
test_data /= std
from keras import models
from keras import layers
model = models.Sequential()
model.add(layers.Dense(64, activation='relu',
input_shape=(train_data.shape[1],)))
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(1,activation=None))
model.compile(optimizer='rmsprop', loss='mse', metrics=['mae'])
history = model.fit(train_data, train_targets,epochs=7, batch_size=1, verbose=1)
-----------------------------------------------------------------------------
Epoch 1/7
404/404 [==============================] - 1s 1ms/step - loss: 309.4995 - mae: 14.3059
Epoch 2/7
404/404 [==============================] - 0s 1ms/step - loss: 29.8972 - mae: 3.6916
Epoch 3/7
404/404 [==============================] - 0s 1ms/step - loss: 19.3192 - mae: 3.0052
Epoch 4/7
404/404 [==============================] - 0s 1ms/step - loss: 11.1953 - mae: 2.4848
Epoch 5/7
404/404 [==============================] - 0s 1ms/step - loss: 12.8683 - mae: 2.5286
Epoch 6/7
404/404 [==============================] - 0s 1ms/step - loss: 14.1816 - mae: 2.5489
Epoch 7/7
404/404 [==============================] - 0s 1ms/step - loss: 9.3017 - mae: 2.1364

أهم معايير قياس كفاءة النماذج في كيراس وتنسرفلو:

Keras Regression Metrics:
Mean Squared Error: mean_squared_error, MSE or mse
Mean Absolute Error: mean_absolute_error, MAE, mae
Mean Absolute Percentage Error: mean_absolute_percentage_error, MAPE, mape
Cosine Proximity: cosine_proximity, cosine
Keras Classification Metrics
Binary Accuracy: binary_accuracy, acc
Categorical Accuracy: categorical_accuracy, acc
Sparse Categorical Accuracy: sparse_categorical_accuracy