Accelerated, Parallel and Proximal Coordinate Descent

12/20/2013
by   Olivier Fercoq, et al.
0

We propose a new stochastic coordinate descent method for minimizing the sum of convex functions each of which depends on a small number of coordinates only. Our method (APPROX) is simultaneously Accelerated, Parallel and PROXimal; this is the first time such a method is proposed. In the special case when the number of processors is equal to the number of coordinates, the method converges at the rate 2ω̅L̅ R^2/(k+1)^2 , where k is the iteration counter, ω̅ is an average degree of separability of the loss function, L̅ is the average of Lipschitz constants associated with the coordinates and individual functions in the sum, and R is the distance of the initial point from the minimizer. We show that the method can be implemented without the need to perform full-dimensional vector operations, which is the major bottleneck of existing accelerated coordinate descent methods. The fact that the method depends on the average degree of separability, and not on the maximum degree of separability, can be attributed to the use of new safe large stepsizes, leading to improved expected separable overapproximation (ESO). These are of independent interest and can be utilized in all existing parallel stochastic coordinate descent algorithms based on the concept of ESO.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
09/23/2013

Smooth minimization of nonsmooth functions with parallel coordinate descent methods

We study the performance of a family of randomized parallel coordinate d...
research
12/04/2012

Parallel Coordinate Descent Methods for Big Data Optimization

In this work we show that randomized (block) coordinate descent methods ...
research
10/08/2013

Distributed Coordinate Descent Method for Learning with Big Data

In this paper we develop and analyze Hydra: HYbriD cooRdinAte descent me...
research
01/27/2019

99

It is well known that many optimization methods, including SGD, SAGA, an...
research
02/11/2020

Variance Reduced Coordinate Descent with Acceleration: New Method With a Surprising Application to Finite-Sum Problems

We propose an accelerated version of stochastic variance reduced coordin...
research
12/08/2017

Stochastic Dual Coordinate Descent with Bandit Sampling

Coordinate descent methods minimize a cost function by updating a single...
research
10/07/2013

Parallel coordinate descent for the Adaboost problem

We design a randomised parallel version of Adaboost based on previous st...

Please sign up or login with your details

Forgot password? Click here to reset