Working to improve the work of "Hieu Nguyen and Richard Tsai" in https://arxiv.org/pdf/2107.13184.pdf by seperating non-linear and linear operations with respect to the underlying physics. Goal: removal of ghost effects from paper.
See mnist.py for a test use case of this 'spatially variant' convolution. See Heatmaps/ for test outputs during training of the novel part of this convolutional layer.