Dynamic Feature Acquisition with Arbitrary Conditional Flows
This addresses the challenge of feature acquisition in real-world ML applications where data is uncertain, offering a method to balance prediction improvement with acquisition costs.
The paper tackles the problem of dynamically acquiring additional features to improve predictions under limited data, using conditional mutual information to select informative features and arbitrary conditional flows for distribution estimation, achieving superior performance over baselines in multiple settings.
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.