Convex Optimization with Nonconvex Oracles

نویسندگان

  • Oren Mangoubi
  • Nisheeth K. Vishnoi
چکیده

In machine learning and optimization, one often wants to minimize a convex objective function F but can only evaluate a noisy approximation F̂ to it. Even though F is convex, the noise may render F̂ nonconvex, making the task of minimizing F intractable in general. As a consequence, several works in theoretical computer science, machine learning and optimization have focused on coming up with polynomial time algorithms to minimize F under conditions on the noise F (x) − F̂ (x) such as its uniform-boundedness, or on F such as strong convexity. However, in many applications of interest these conditions do not hold. Here we show that, if the noise has magnitude αF (x) + β for some α, β > 0, then there is a polynomial time algorithm to find an approximate minimizer of F . In particular, our result allows for unbounded noise and generalizes those of [1,17] who proved similar results for the bounded noise case, and that of [2] who assume that the noise grows in a very specific manner and that F is strongly convex. Turning our result on its head, one may also view our algorithm as minimizing a nonconvex function F̂ that is promised to be related to a convex function F as above. Technically, Markov chains, such as the stochastic gradient Langevin dynamics, are deployed to arrive at approximate solutions to these optimization problems. For the class of noise we consider, no single temperature allows such a Markov chain to both mix quickly and concentrate near the global minimizer. Consequently, our algorithm, which is a “simulated annealing” modification of the stochastic gradient Langevin dynamics, gradually decreases the temperature of the chain to approach the global minimizer. Analyzing such an algorithm for the unbounded noise model and a general convex function turns out to be challenging and requires several technical ideas that might be of independent interest in deriving non-asymptotic bounds for other simulated annealing based algorithms. 1 ar X iv :1 71 1. 02 62 1v 1 [ cs .D S] 7 N ov 2 01 7

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Linear Time Varying MPC Based Path Planning of an Autonomous Vehicle via Convex Optimization

In this paper a new method is introduced for path planning of an autonomous vehicle. In this method, the environment is considered cluttered and with some uncertainty sources. Thus, the state of detected object should be estimated using an optimal filter. To do so, the state distribution is assumed Gaussian. Thus the state vector is estimated by a Kalman filter at each time step. The estimation...

متن کامل

Inexactness in Bundle Methods for Locally Lipschitz Functions

We consider the problem of computing a critical point of a nonconvex locally Lipschitz function over a convex compact constraint set given an inexact oracle that provides an approximate function value and an approximate subgradient. We assume that the errors in function and subgradient evaluations are merely bounded, and in particular need not vanish in the limit. After some discussion on how t...

متن کامل

Parallel Asynchronous Stochastic Variance Reduction for Nonconvex Optimization

Nowadays, asynchronous parallel algorithms have received much attention in the optimization field due to the crucial demands for modern large-scale optimization problems. However, most asynchronous algorithms focus on convex problems. Analysis on nonconvex problems is lacking. For the Asynchronous Stochastic Descent (ASGD) algorithm, the best result from (Lian et al., 2015) can only achieve an ...

متن کامل

Convex Relaxation Methods for Nonconvex Polynomial Optimization Problems

This paper introduces to constructing problems of convex relaxations for nonconvex polynomial optimization problems. Branch-and-bound algorithms are convex relaxation based. The convex envelopes are of primary importance since they represent the uniformly best convex underestimators for nonconvex polynomials over some region. The reformulationlinearization technique (RLT) generates LP (linear p...

متن کامل

Information-theoretic lower bounds for convex optimization with erroneous oracles

We consider the problem of optimizing convex and concave functions with access to an erroneous zeroth-order oracle. In particular, for a given function x → f(x) we consider optimization when one is given access to absolute error oracles that return values in [f(x) − , f(x) + ] or relative error oracles that return value in [(1− )f(x), (1 + )f(x)], for some > 0. We show stark information theoret...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:
  • CoRR

دوره abs/1711.02621  شماره 

صفحات  -

تاریخ انتشار 2017