I have gone through the official documentation but still can't understand what actually TimeDistributed does as a layer in Keras model?
I couldn't understand the difference between TimeDistributed and TimeDistributedDense? When will someone use TimeDistributedDense? Is it only to reduce training data set? Does it have other benefit?
Can anyone explain with a precise example that what these two type of layer wrappers does?
So - basically the TimeDistributedDense was introduced first in early versions of Keras in order to apply a Dense layer stepwise to sequences. TimeDistributed is a Keras wrapper which makes possible to get any static (non-sequential) layer and apply it in a sequential manner. So if e.g. your layer accepts as an input something of shape (d1, .., dn) thanks to TimeDistributed wrapper your layer could accept an input with a shape of (sequence_len, d1, ..., dn) by applying a layer provided to X[0,:,:,..,:], X[1,:,...,:], ..., X[len_of_sequence,:,...,:].
An example of such usage might be using a e.g. pretrained convolutional layer to a short video clip by applying TimeDistributed(conv_layer) where conv_layer is applied to each frame of a clip. It produces the sequence of outputs which might be then consumed by next recurrent or TimeDistributed layer.
It's good to know that usage of TimeDistributedDense is depreciated and it's better to use TimeDistributed(Dense).
TimeDistributedDense is the same than TimeDistributed with the only difference that TimeDistributed can be used with different types of layer and not just Dense layer.
Keras documentation says that about TimeDistributed :
"Note this is strictly equivalent to using layers.core.TimeDistributedDense. However what is different about TimeDistributed is that it can be used with arbitrary layers, not just Dense, for instance with a Convolution2D layer"
