SpAM: Sparse Additive Models
Pradeep Ravikumar, Han Liu, John Lafferty, Larry Wasserman Carnegie Mellon University

T64

We present methods and theor y for functional sparsity in high dimensional nonparametric additive models: Y= Main results: · Formulation of convex optimization problem · Sparse backfitting algorithm · Theoretical results on persistence (risk consistency), sparsistency (consistency of sparsity pattern), L2 consistency jp
=1

fj (Xj ) +

p

n, most fj  0