GMMILoss

Differentiable Mutual information Loss with pytorch

1. Train a GM
2. Estimate MI using covarience matrix (easy)
3. Consruct a loss function with estimated covarience matrix and means
4. BackPropagate!

Todo:

• update code
• works quite bad with High dimensional data, check this paper: https://openreview.net/forum?id=SvrYl-FDq2