Here is a model that trains time series data in (batch, step, features) way.
I have kept the random state for train test split function the same. Every parameter below the same, running the model training yields different outcomes every time and the outcomes are drastically different.
What may be the factors that led to this? Regularization?
X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.2, random_state=666)
def attention_model(X_train, y_train, X_test, y_test,num_classes,dropout=0.2, batch_size=68, learning_rate=0.0001,epochs=20,optimizer='Adam'):
Dense_unit = 12
LSTM_unit = 12
attention_param = LSTM_unit*2
attention_init_value = 1.0/attention_param
u_train = np.full((X_train.shape[0], attention_param),
attention_init_value, dtype=np.float32)
u_test = np.full((X_test.shape[0],attention_param),
attention_init_value, dtype=np.float32)
with keras.backend.name_scope('BLSTMLayer'):
# Bi-directional Long Short-Term Memory for learning the temporal aggregation
input_feature = Input(shape=(X_train.shape[1],X_train.shape[2]))
x = Masking(mask_value=0)(input_feature)
x = Dense(Dense_unit,kernel_regularizer=l2(0.005), activation='relu')(x)
x = Dropout(dropout)(x)
x = Dense(Dense_unit,kernel_regularizer=l2(0.005),activation='relu')(x)
x = Dropout(dropout)(x)
x = Dense(Dense_unit,kernel_regularizer=l2(0.005),activation='relu')(x)
x = Dropout(dropout)(x)
x = Dense(Dense_unit,kernel_regularizer=l2(0.005), activation='relu')(x)
x = Dropout(dropout)(x)
y = Bidirectional(LSTM(LSTM_unit,activity_regularizer=l2(0.000029),kernel_regularizer=l2(0.027),recurrent_regularizer=l2(0.025),return_sequences=True, dropout=dropout))(x)
# y = Bidirectional(LSTM(LSTM_unit, kernel_regularizer=l2(0.01),recurrent_regularizer=l2(0.01), return_sequences=True, dropout=dropout))(y)
with keras.backend.name_scope('AttentionLayer'):
# Logistic regression for learning the attention parameters with a standalone feature as input
input_attention = Input(shape=(LSTM_unit * 2,))
u = Dense(LSTM_unit * 2, activation='softmax')(input_attention)
# To compute the final weights for the frames which sum to unity
alpha = dot([u, y], axes=-1) # inner prod.
alpha = Activation('softmax')(alpha)
with keras.backend.name_scope('WeightedPooling'):
# Weighted pooling to get the utterance-level representation
z = dot([alpha, y], axes=1)
# Get posterior probability for each emotional class
output = Dense(num_classes, activation='softmax')(z)
model = Model(inputs=[input_attention, input_feature], outputs=output)
optimizer = opt_select(optimizer,learning_rate)
model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer=optimizer)
hist = model.fit([u_train, X_train],
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=2,
validation_data=([u_test, X_test], y_test))
#kernel_regularizer=l2(0.002),recurrent_regularizer=l2(0.002),
return hist
batch_size= 150
#217
epochs = 1000
learning_rate = 0.00081
optimizer = 'RMS'
num_classes = y_train.shape[1]
dropout=0.22
tf.keras.backend.clear_session()
history = attention_model(X_train, y_train, X_test, y_test, num_classes,dropout = dropout,batch_size=batch_size, learning_rate=learning_rate,epochs=epochs,optimizer=optimizer
)
