Dynamic Feature Acquisition with Arbitrary Conditional Flows

TL;DR

This paper introduces a dynamic feature acquisition framework using arbitrary conditional flows and Bayesian networks, enabling models to selectively acquire features based on information gain, improving prediction accuracy efficiently.

Contribution

It proposes a novel method combining ACFlow and Bayesian networks for dynamic feature acquisition guided by conditional mutual information.

Findings

Outperforms baseline methods in various settings

Efficiently balances acquisition cost and prediction accuracy

Demonstrates superior adaptability in real-world scenarios

Abstract

Many real-world situations allow for the acquisition of additional relevant information when making an assessment with limited or uncertain data. However, traditional ML approaches either require all features to be acquired beforehand or regard part of them as missing data that cannot be acquired. In this work, we propose models that dynamically acquire new features to further improve the prediction assessment. To trade off the improvement with the cost of acquisition, we leverage an information theoretic metric, conditional mutual information, to select the most informative feature to acquire. We leverage a generative model, arbitrary conditional flow (ACFlow), to learn the arbitrary conditional distributions required for estimating the information metric. We also learn a Bayesian network to accelerate the acquisition process. Our model demonstrates superior performance over baselines evaluated in multiple settings.