Classification with Costly Features as a Sequential Decision-Making Problem

TL;DR

This paper formulates the classification with costly features as a sequential decision-making problem, using deep reinforcement learning to optimize feature acquisition under budget constraints, and demonstrates robust performance across multiple datasets.

Contribution

It introduces an explicit MDP formulation for cost-sensitive classification and applies deep RL to solve it, handling real-world scenarios with limited data and missing features.

Findings

Outperforms prior algorithms across various datasets

Robustly handles limited and missing feature data

Flexible framework adaptable with domain-independent RL improvements

Abstract

This work focuses on a specific classification problem, where the information about a sample is not readily available, but has to be acquired for a cost, and there is a per-sample budget. Inspired by real-world use-cases, we analyze average and hard variations of a directly specified budget. We postulate the problem in its explicit formulation and then convert it into an equivalent MDP, that can be solved with deep reinforcement learning. Also, we evaluate a real-world inspired setting with sparse training dataset with missing features. The presented method performs robustly well in all settings across several distinct datasets, outperforming other prior-art algorithms. The method is flexible, as showcased with all mentioned modifications and can be improved with any domain independent advancement in RL.