Data Analytics for Business and Societal Challenges

Loading...

Media is loading
 

Paper Number

1939

Paper Type

short

Description

Multi-armed bandit problem is a widely used framework for sequential decision making under uncertainty. In its standard formulation, the bandit problem assumes that the reward distributions are stationary and independent. These assumptions are not satisfied in many practical settings such as financial portfolio selection where are rewards are correlated and are subjected to seasonality, external shocks, etc. In this paper, we propose an algorithm to leverage the dependence structure between the rewards, through the use of copulas, in non-stationary settings. Our preliminary experiments on simulated and real data demonstrate the advantage of copula-based modeling of rewards, where the performance of our approach is superior to state-of-the-art non-stationary MAB algorithms.

Comments

14-Data

Download

Share

COinS
 
Dec 12th, 12:00 AM

Dependency Modeling with Copulas in Multi-Armed Bandits

Multi-armed bandit problem is a widely used framework for sequential decision making under uncertainty. In its standard formulation, the bandit problem assumes that the reward distributions are stationary and independent. These assumptions are not satisfied in many practical settings such as financial portfolio selection where are rewards are correlated and are subjected to seasonality, external shocks, etc. In this paper, we propose an algorithm to leverage the dependence structure between the rewards, through the use of copulas, in non-stationary settings. Our preliminary experiments on simulated and real data demonstrate the advantage of copula-based modeling of rewards, where the performance of our approach is superior to state-of-the-art non-stationary MAB algorithms.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.