Getting Ed's agreement on need for sparsity-inducing prior

[hyunjimoon]

From the paper, Combining stock-and-flow, agent-based, and social network methods to model team performance, Ed commented we could start from large enough N by N matrix and aim for sparsity.

I think horse-shoe prior might be the first to look into although I had a feeling Michael's writing shows evidence against that prior. Also, lasso i.e. L0 regularization does not always filter useful features.

[hyunjimoon]

Relevant to #69