I am trying to create a autoencoder for anomally detection using tensorflow, the model runs but thats about as far as it goes. Here is the model:
int_vectorizer = layers.TextVectorization( max_tokens=10000, output_mode='int', output_sequence_length=140 ) int_vectorizer.adapt(adapt_data) model = tf.keras.Sequential([ int_vectorizer, layers.Dense(256, activation="relu"), layers.Dense(128, activation="relu"), layers.Dense(64, activation="relu"), layers.Dense(32, activation="relu"), layers.Dense(64, activation="relu"), layers.Dense(128, activation="sigmoid"),` ]) label_converter = layers.StringLookup(output_mode="int") label_converter.adapt(data) model.compile(optimizer='adam', loss="mae") model.fit(data, label_converter(data), epochs=200) When testing it, i get the same percentage levels for the following 2 inputs:
I tried changing layers, amounts of nodes, amounts of units and different activations.
Источник: https://stackoverflow.com/questions/781 ... oder-model
