I’m quite excited by the Nuit Blanche post on the papers “Structure Discovery in Nonparametric Regression through Compositional Kernel Search” (Duvenaudy, Lloydy, Grossez, Tenenbaumz, Ghahramaniy 2013) and “Exploiting compositionality to explore a large space of model structures” (Grosse, Salakhutdinovm, Freeman, and Tenenbaum 2012). For years my old company Momentum Investment Services, Carl, and I have been looking for fast, systematic ways to search large hypothesis spaces. We considered context-free grammars as a means of generating hypothesis. Carl and I did not get anywhere with that research, but now it seems that others have succeeded. Be sure to look over the article, the blog posts, and the comments.