25 citations · 77 across the 32 of their papers we have counts for
7 papers · 1 filter
Global Convergence Rate Analysis of a Generic Line Search Algorithm with Noise
Albert S. Berahas, Liyuan Cao, Katya Scheinberg
In this paper, we develop convergence analysis of a modified line search method for objective functions whose value is computed with noise and whose gradient estimates are inexact…
Linear interpolation gives better gradients than Gaussian smoothing in derivative-free optimization
Albert S Berahas, Liyuan Cao, Krzysztof Choromanski +1
In this paper, we consider derivative free optimization problems, where the objective function is smooth but is computed with some amount of noise, the function evaluations are exp…
Scaling Up Quasi-Newton Algorithms: Communication Efficient Distributed SR1
Majid Jahani, Mohammadreza Nazari, Sergey Rusakov +2
In this paper, we present a scalable distributed implementation of the Sampled Limited-memory Symmetric Rank-1 (S-LSR1) algorithm. First, we show that a naive distributed implement…
A Theoretical and Empirical Comparison of Gradient Approximations in Derivative-Free Optimization
Albert S. Berahas, Liyuan Cao, Krzysztof Choromanski +1
In this paper, we analyze several methods for approximating gradients of noisy functions using only function values. These methods include finite differences, linear interpolation,…
Nested Distributed Gradient Methods with Adaptive Quantized Communication
Albert S. Berahas, Charikleia Iakovidou, Ermin Wei
In this paper, we consider minimizing a sum of local convex objective functions in a distributed setting, where communication can be costly. We propose and analyze a class of neste…
Limited-Memory BFGS with Displacement Aggregation
Albert S. Berahas, Frank E. Curtis, Baoyu Zhou
A displacement aggregation strategy is proposed for the curvature pairs stored in a limited-memory BFGS (a.k.a. L-BFGS) method such that the resulting (inverse) Hessian approximati…
Quasi-Newton Methods for Machine Learning: Forget the Past, Just Sample
Albert S. Berahas, Majid Jahani, Peter Richtárik +1
We present two sampled quasi-Newton methods (sampled LBFGS and sampled LSR1) for solving empirical risk minimization problems that arise in machine learning. Contrary to the classi…