jacob_drori comments on Do sparse autoencoders find “true features”?

[jacob_drori]

Nice, this is exactly what I was asking for. Thanks!