Keras TimeDistributed - total weight?
From keras docs : Then you can use TimeDistributed
to apply a layer Dense
to each of the ten timestamps independently:
# as the first layer in a model
model = Sequential()
model.add(TimeDistributed(Dense(8), input_shape=(10, 16)))
# now model.output_shape == (None, 10, 8)
# subsequent layers: no need for input_shape
model.add(TimeDistributed(Dense(32)))
# now model.output_shape == (None, 10, 32)
I can't find it anywhere, Are the layer weights Dense
separated along the time axis?
Yes, they are separate - timestep
the same applies for each Dense
. What's more - a Keras 2.0
behavior like this TimeDistributed
is now the default for a layer Dense
applied to an input that is more than 2D (including batch_dimension
).