Not easy to see what's going on in the plots provided by Pylearn2. Might as well use a grid of heatmaps in holoviews to get a better picture. Can also extend to looking at the kernels of all convolutional layer. Will probably require putting some code into the tools repo.
Should also add the function for plotting arbitrary collections of channels into the module in tools that will be created.
Not easy to see what's going on in the plots provided by Pylearn2. Might as well use a grid of heatmaps in holoviews to get a better picture. Can also extend to looking at the kernels of all convolutional layer. Will probably require putting some code into the tools repo.
Should also add the function for plotting arbitrary collections of channels into the module in tools that will be created.