# Bandits attack function optimization

Source: [Bandits attack function optimization](<https://www.sota2.com/research/paper/69facb3dc0f56ed84308a57e>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-07T05:00:04.653Z

Authors: Philippe Preux, R\\'emi Munos, Michal Valko

Published: 2026-05-05T08:33:44.000Z

## Abstract

We consider function optimization as a sequential decision making problem under budget constraint. This constraint limits the number of objective function evaluations allowed during the optimization. We consider an algorithm inspired by a continuous version of a multi-armed bandit problem which attacks this optimization problem by solving the tradeoff between exploration (initial quasi-uniform search of the domain) and exploitation (local optimization around the potentially global maxima). We introduce the so-called Simultaneous Optimistic Optimization (SOO), a deterministic algorithm that works by domain partitioning. The benefit of such approach are the guarantees on the returned solution and the numerical efficiency of the algorithm. We present this machine learning approach to optimization, and provide the empirical assessment of SOO on the CEC'2014 competition on single objective real-parameter numerical optimization test-suite.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Global Optimization on CEC 10D 2014 (30 Functions)](<https://www.sota2.com/research/sota/global-optimization-on-cec-10d-2014-30-functions.md>): Global Optimization; CEC 10D 2014; Objective Function Value (F1): 1139.36.

- [Global Optimization on CEC 30D 2014](<https://www.sota2.com/research/sota/global-optimization-on-cec-30d-2014.md>): Global Optimization; CEC 30D 2014; F1 Value: 502760.

- [Single objective real parameter numerical optimization on CEC 10 Dimensions 2014 (F1-F30)](<https://www.sota2.com/research/sota/single-objective-real-parameter-numerical-optimization-on-cec-10-dimensions-2014-f1-f30.md>): Single objective real parameter numerical optimization; CEC 10 Dimensions 2014; Objective Value (F1): 1139.36.

- [Single objective real parameter numerical optimization on CEC 30 Dimensions 2014 (F1-F30)](<https://www.sota2.com/research/sota/single-objective-real-parameter-numerical-optimization-on-cec-30-dimensions-2014-f1-f30.md>): Single objective real parameter numerical optimization; CEC 30 Dimensions 2014; F1 Objective Value: 319606.

- [Single objective real parameter numerical optimization on CEC 50 Dimensions 2014 (F1-F30)](<https://www.sota2.com/research/sota/single-objective-real-parameter-numerical-optimization-on-cec-50-dimensions-2014-f1-f30.md>): Single objective real parameter numerical optimization; CEC 50 Dimensions 2014; Objective Value (F1): 400626.

- [Global Optimization on CEC 50D 2014 (30 Functions)](<https://www.sota2.com/research/sota/global-optimization-on-cec-50d-2014-30-functions.md>): Global Optimization; CEC 50D 2014; Objective Value (F1): 500290.

- [Global Optimization on CEC 100D 2014 (30 Functions)](<https://www.sota2.com/research/sota/global-optimization-on-cec-100d-2014-30-functions.md>): Global Optimization; CEC 100D 2014; F1 Value: 2302870.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.03496>)

[PDF](<https://arxiv.org/pdf/2605.03496.pdf>)