One issue is that ICA covers a few different implementations. My favourite would be Infomax '95 paper by Bell and Sejnowski. Under this setup: ICA is a single

One issue is that ICA covers a few different implementations. My favourite would be Infomax '95 paper by Bell and Sejnowski. Under this setup: ICA is a single layer neural network that maps N inputs to N outputs. The loss function being maximised is the mutual info between input and output.