Conditional Information Gain Networks

Ufuk Can Biçici, Cem Keskin, Lale Akarun

Deep neural network models owe their representational power to the high number of learnable parameters. It is often infeasible to run these largely parametrized deep models in limi…