A Contextual Combinatorial Semi-Bandit Approach to Network Bottleneck Identification

TL;DR

This paper introduces a unified online learning framework using combinatorial semi-bandits for network bottleneck identification, capable of leveraging contextual information and evaluated on real-world road networks.

Contribution

It develops a novel framework that combines various semi-bandit algorithms with contextual information for efficient bottleneck detection in incomplete networks.

Findings

Effective bottleneck identification in real-world road networks

Framework adapts multiple semi-bandit algorithms including NeuralUCB and Thompson Sampling

Demonstrates robustness across different network settings

Abstract

Bottleneck identification is a challenging task in network analysis, especially when the network is not fully specified. To address this task, we develop a unified online learning framework based on combinatorial semi-bandits that performs bottleneck identification in parallel with learning the specifications of the underlying network. Within this framework, we adapt and study various combinatorial semi-bandit methods such as epsilon-greedy, LinUCB, BayesUCB, NeuralUCB, and Thompson Sampling. In addition, our framework is capable of using contextual information in the form of contextual bandits. Finally, we evaluate our framework on the real-world application of road networks and demonstrate its effectiveness in different settings.