Duration: 53:56

489 views

1

Optimization problems with sparsity constraints have attracted much attention in recent years. There are two types of methods: convex relaxation such as L1 regularization and greedy algorithms. Although the former has received more attention in the machine learning community, my opinion is that the latter approach is more flexible and powerful. This talk will present various greedy and local search algorithms for sparsity constrained optimization, focusing on their theoretical properties.