Approximate Maximum A Posteriori Inference with Entropic Priors

Matthew D. Hoffman

In certain applications it is useful to fit multinomial distributions to observed data with a penalty term that encourages sparsity. For example, in probabilistic latent audio sour…