c4b8bb990423f770dd7f26ff79168416-Paper.pdf

In machine learning, it is often valuable for models to be parsimonious orsparsefor a variety of reasons, from memory savings and computational speedups tomodel interpretability and generalizability.