Learning of Generalized Low-Rank Models: A Greedy Approach
This is an incremental improvement for machine learning practitioners working on low-rank matrix problems, as it extends applicability to more objective functions while maintaining efficiency.
The paper tackled the limitation of existing low-rank matrix learning algorithms like R1MP, which only work with square loss, by developing a more flexible greedy algorithm for generalized low-rank models that handles various objective types, resulting in much faster speed with comparable or better prediction performance.
Learning of low-rank matrices is fundamental to many machine learning applications. A state-of-the-art algorithm is the rank-one matrix pursuit (R1MP). However, it can only be used in matrix completion problems with the square loss. In this paper, we develop a more flexible greedy algorithm for generalized low-rank models whose optimization objective can be smooth or nonsmooth, general convex or strongly convex. The proposed algorithm has low per-iteration time complexity and fast convergence rate. Experimental results show that it is much faster than the state-of-the-art, with comparable or even better prediction performance.