Timezone: »

A Variant of Anderson Mixing with Minimal Memory Size
Fuchao Wei · Chenglong Bao · Yang Liu · Guangwen Yang


Anderson mixing (AM) is a useful method that can accelerate fixed-point iterations by exploring the information from historical iterations. Despite its numerical success in various applications, the memory requirement in AM remains a bottleneck when solving large-scale optimization problems in a resource-limited machine. To address this problem, we propose a novel variant of AM method, called Min-AM, by storing only one vector pair, that is the minimal memory size requirement in AM. Our method forms a symmetric approximation to the inverse Hessian matrix and is proved to be equivalent to the full-memory Type-I AM for solving strongly convex quadratic optimization. Moreover, for general nonlinear optimization problems, we establish the convergence properties of Min-AM under reasonable assumptions and show that the mixing parameters can be adaptively chosen by estimating the eigenvalues of the Hessian. Finally, we extend Min-AM to solve stochastic programming problems. Experimental results on logistic regression and network training problems validate the effectiveness of the proposed Min-AM.

Author Information

Fuchao Wei (Tsinghua University, Tsinghua University)
Chenglong Bao (Tsinghua university)
Yang Liu (Tsinghua University)
Guangwen Yang (Tsinghua University)

More from the Same Authors