ABSTRACT
Bayesian learning provides a general framework for sparse estimation, which includes nonBayesian methods as special cases-the trace-norm regularization and the 1-norm regularization can be interpreted as maximum a posteriori (MAP) estimation with certain prior distributions.
