Sparse Coding and Information in Hebbian Neural Networks

The analysis of the information provided by a neural network with a generalized Hebbian rule shows that after an optimization process, in the limit of very low levels of activity this learning rule is very efficient.