Layers
In this section we will go over the layers that were implemented in this repository.
Bayesian Module
This model implements variational layers:
- Dense Variational layer
FFT Module
This module implements various Discrete Cosine Transform layers in tensorflow, namely:
- Discrete Cosine Transform 3-Dimensional
- Inverse Discrete Cosine Transform 3-Dimensional
- Padded Inverse Discrete Cosine Transform 3-Dimensional
- Variational Inverse Discrete Cosine Transform 3-Dimensional
Random Fourier Module
This module is a wrap of the implementation done in tensorflow, but some tweaks were needed to get it to work with our methodology. The file is similar to the tensorflow implementation.
Topographical Attention Module
This module implements the topographical attention presented in this paper.
Resnet Block Module
This module allows an easy integration of neural architecture specifications that were generated automatically, according to the methodology described in this paper.