My apologies. I gave a full day just to understand that the output of bottleneck layer (the last layer in the Encoder container) is actually a COMPRESSED feature set that somehow tries to match the distribution of the original feature set. Thus, to sum up-
If you originally have 1000 features set, and you want output features to be around 100, then you will get this by using Autoencoder. However, the output 100 features would not be a subset of 1000 input features. They will be all new, compressed form of the initial, 1000 input features.
I am new to Pytorch. I want to know when we train an autoencoder, what are the outputs of the ENCODER container of it?
As I know the Autoencoder is used for Dimension reduction. So, after the encoder part gets executed, do we get a set of indices that correspond to relevant columns from the dataset, or do we get something entirely different?
I am actually using an Autoencoder for feature extraction, and this knowledge will help me quite a lot. I need to extract features from the ENCODER container and I don’t know how to do it.
Thanks in advance.