Блог

Я работаю с LSTM сетью для прогнозирования временных рядов. Я использую keras tuner for hyperparameters tuning , но я всегда получаю нулевую точность, как показано в приведенном ниже коде . Любая помощь была бы отличной.

 tuner_H = Hyperband(
    build_model,
    max_epochs=20,
    objective='val_accuracy',
    seed=1,
    executions_per_trial=10,
    directory='hyperband',
    project_name='cifar10'
)
  tuner_H.search(X_train, Y_train, epochs= 20, validation_split=0.1, verbose = 1)

Epoch 1/3
283/283 [==============================] - 2s 5ms/step - loss: 0.1571 - accuracy: 0.0000e 00 - val_loss: 0.2332 - val_accuracy: 0.0000e 00
Epoch 2/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1502 - accuracy: 0.0000e 00 - val_loss: 0.2236 - val_accuracy: 0.0000e 00
Epoch 3/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1439 - accuracy: 0.0000e 00 - val_loss: 0.2142 - val_accuracy: 0.0000e 00
Epoch 1/3
283/283 [==============================] - 2s 5ms/step - loss: 0.1497 - accuracy: 0.0000e 00 - val_loss: 0.2217 - val_accuracy: 0.0000e 00
Epoch 2/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1427 - accuracy: 0.0000e 00 - val_loss: 0.2116 - val_accuracy: 0.0000e 00
Epoch 3/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1356 - accuracy: 0.0000e 00 - val_loss: 0.2010 - val_accuracy: 0.0000e 00
Epoch 1/3
283/283 [==============================] - 2s 5ms/step - loss: 0.1490 - accuracy: 0.0000e 00 - val_loss: 0.2201 - val_accuracy: 0.0000e 00
Epoch 2/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1418 - accuracy: 1.1058e-04 - val_loss: 0.2098 - val_accuracy: 0.0000e 00
Epoch 3/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1352 - accuracy: 0.0000e 00 - val_loss: 0.1996 - val_accuracy: 0.0000e 00
Epoch 1/3
283/283 [==============================] - 2s 5ms/step - loss: 0.1472 - accuracy: 0.0000e 00 - val_loss: 0.2172 - val_accuracy: 0.0000e 00
Epoch 2/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1402 - accuracy: 0.0000e 00 - val_loss: 0.2066 - val_accuracy: 0.0000e 00
Epoch 3/3
283/283 [==============================] - 1s 4ms/step - loss: 0.1332 - accuracy: 0.0000e 00 - val_loss: 0.1960 - val_accuracy: 0.0000e 00.........................