Block stochastic gradient iteration for convex and nonconvex optimization
This method addresses optimization challenges in machine learning and data analysis by reducing per-iteration costs for large-scale problems, though it is incremental as it builds on existing SG and BCD techniques.
The authors tackled the problem of optimizing objective functions with many components and multiple variable blocks by introducing a block stochastic gradient (BSG) method that combines stochastic gradient and block coordinate descent. They established convergence rates matching SG for convex cases and proved convergence for nonconvex cases, with numerical tests on problems like stochastic least squares and low-rank tensor recovery.
The stochastic gradient (SG) method can minimize an objective function composed of a large number of differentiable functions, or solve a stochastic optimization problem, to a moderate accuracy. The block coordinate descent/update (BCD) method, on the other hand, handles problems with multiple blocks of variables by updating them one at a time; when the blocks of variables are easier to update individually than together, BCD has a lower per-iteration cost. This paper introduces a method that combines the features of SG and BCD for problems with many components in the objective and with multiple (blocks of) variables. Specifically, a block stochastic gradient (BSG) method is proposed for solving both convex and nonconvex programs. At each iteration, BSG approximates the gradient of the differentiable part of the objective by randomly sampling a small set of data or sampling a few functions from the sum term in the objective, and then, using those samples, it updates all the blocks of variables in either a deterministic or a randomly shuffled order. Its convergence for both convex and nonconvex cases are established in different senses. In the convex case, the proposed method has the same order of convergence rate as the SG method. In the nonconvex case, its convergence is established in terms of the expected violation of a first-order optimality condition. The proposed method was numerically tested on problems including stochastic least squares and logistic regression, which are convex, as well as low-rank tensor recovery and bilinear logistic regression, which are nonconvex.