Dynamic Ensemble Active Learning: A Non-Stationary Bandit with Expert Advice

TL;DR

This paper introduces a dynamic ensemble active learning method using a non-stationary bandit approach with expert advice, which adaptively selects the best criteria during active learning sessions, improving performance across multiple datasets.

Contribution

It proposes a novel dynamic ensemble active learning framework based on non-stationary bandits with expert advice, addressing the non-stationarity of criteria effectiveness within sessions.

Findings

The method has theoretical guarantees.

It performs well on 13 popular datasets.

It adaptively selects criteria during active learning.

Abstract

Active learning aims to reduce annotation cost by predicting which samples are useful for a human teacher to label. However it has become clear there is no best active learning algorithm. Inspired by various philosophies about what constitutes a good criteria, different algorithms perform well on different datasets. This has motivated research into ensembles of active learners that learn what constitutes a good criteria in a given scenario, typically via multi-armed bandit algorithms. Though algorithm ensembles can lead to better results, they overlook the fact that not only does algorithm efficacy vary across datasets, but also during a single active learning session. That is, the best criteria is non-stationary. This breaks existing algorithms' guarantees and hampers their performance in practice. In this paper, we propose dynamic ensemble active learning as a more general and promising research direction. We develop a dynamic ensemble active learner based on a non-stationary multi-armed bandit with expert advice algorithm. Our dynamic ensemble selects the right criteria at each step of active learning. It has theoretical guarantees, and shows encouraging results on $13$ popular datasets.