Closed JacobGlennAyers closed 1 year ago
We can leave things in terms of the convex dual and run the evolutionary algorithms on the number of sign patterns necessary. I do not know how to convert the convex cnn to primal. Although the benefit is we get interpretability as we can explain how each feature contributes to identifying and denoising the image
It seems that scnn already has an implementation of this that we might use