On Convergence of Distributed Approximate Newton Methods: Globalization, Sharper Bounds and Beyond
This work addresses convergence challenges in communication-efficient distributed optimization, offering incremental improvements for researchers and practitioners in machine learning.
The paper tackles the convergence issues of the DANE algorithm for distributed machine learning by proposing new variants with backtracking line search and heavy-ball acceleration, achieving global asymptotic convergence and sharper local non-asymptotic rates for strongly convex functions, with numerical evidence supporting theoretical advantages.
The DANE algorithm is an approximate Newton method popularly used for communication-efficient distributed machine learning. Reasons for the interest in DANE include scalability and versatility. Convergence of DANE, however, can be tricky; its appealing convergence rate is only rigorous for quadratic objective, and for more general convex functions the known results are no stronger than those of the classic first-order methods. To remedy these drawbacks, we propose in this paper some new alternatives of DANE which are more suitable for analysis. We first introduce a simple variant of DANE equipped with backtracking line search, for which global asymptotic convergence and sharper local non-asymptotic convergence rate guarantees can be proved for both quadratic and non-quadratic strongly convex functions. Then we propose a heavy-ball method to accelerate the convergence of DANE, showing that nearly tight local rate of convergence can be established for strongly convex functions, and with proper modification of algorithm the same result applies globally to linear prediction models. Numerical evidence is provided to confirm the theoretical and practical advantages of our methods.