Commit dcd20ea4 authored by lucas_miranda's avatar lucas_miranda
Browse files

Implemented VAEP hypermodel in hypermodels.py

parent b9f7567a
This diff is collapsed.
...@@ -531,7 +531,10 @@ class SEQ_2_SEQ_MMVAE: ...@@ -531,7 +531,10 @@ class SEQ_2_SEQ_MMVAE:
pass pass
# TODO next: # TODO:
# - MERGE BatchNormalization layers in generator and _generator in SEQ_2_SEQ_VAE # - cleaner implementation of reparameterization trick (sampling on input, outside the main graph)
# - VAE loss function (though this should be analysed later on taking the encodings into account) # - KL / MMD warmup (Ladder Variational Autoencoders)
# - Smaller input sliding window (10-15 frames) # - Gaussian Mixture + Categorical priors -> Deep Clustering
# - free bits paper
# - Attention mechanism for encoder / decoder (does it make sense?)
# - Transformer encoder/decoder (does it make sense?)
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment