Neural Active Learning Meets the Partial Monitoring Framework

TL;DR

This paper introduces a novel theoretical framework for online active learning using partial monitoring, proposing NeuralCBP, a neural network-based strategy that effectively balances information acquisition and prediction costs.

Contribution

It establishes partial monitoring as a foundation for online active learning and introduces NeuralCBP, the first deep neural network-based partial monitoring strategy for this setting.

Findings

NeuralCBP outperforms state-of-the-art baselines on various tasks.

The framework unifies existing OAL tasks under partial monitoring.

NeuralCBP effectively handles cost-sensitive and multi-class scenarios.

Abstract

We focus on the online-based active learning (OAL) setting where an agent operates over a stream of observations and trades-off between the costly acquisition of information (labelled observations) and the cost of prediction errors. We propose a novel foundation for OAL tasks based on partial monitoring, a theoretical framework specialized in online learning from partially informative actions. We show that previously studied binary and multi-class OAL tasks are instances of partial monitoring. We expand the real-world potential of OAL by introducing a new class of cost-sensitive OAL tasks. We propose NeuralCBP, the first PM strategy that accounts for predictive uncertainty with deep neural networks. Our extensive empirical evaluation on open source datasets shows that NeuralCBP has favorable performance against state-of-the-art baselines on multiple binary, multi-class and cost-sensitive OAL tasks.