cs.LGAug 3, 2026

BRiG-AFA: Bellman Risk-to-Go Learning for Non-Myopic Active Feature Acquisition

Authors: Jiaorong FengQian LiYing Li

Organizations: Curtin Business School, Curtin University, Perth, Western Australia, Australia · †Present affiliation: Independent Researcher. · School of Electrical Engineering, Computing and Mathematical Sciences, Curtin University, Perth, Western Australia, Australia

Abstract

Active feature acquisition (AFA) asks which unobserved feature to measure next for each test instance under a budget. Greedy rules are easy to train but can overlook context features whose value is realized only through later acquisitions, while reinforcement-learning and generative approaches introduce difficult optimization or conditional-density estimation. We introduce \method, a deployable, supervised alternative that learns a separate candidate-conditioned risk-to-go function for every remaining budget. Starting from the one-step terminal classification risk, the functions are fitted backward with Bellman targets; inference greedily minimizes the learned terminal risk using only observed values, the mask, candidate identity, and remaining budget. A controlled non-myopic benchmark shows the expected mechanism: at budgets two and three, \method improves accuracy over its one-step ablation by 4.84±2.174.84\pm2.17 and 4.39±1.104.39\pm1.10 percentage points (mean ±\pm standard error over five seeds). On Fashion-MNIST with 20 candidate pixels, it improves accuracy at every nontrivial reported budget on average, including 10.20±0.7410.20\pm0.74 points at four acquisitions; its mean paired gain across budgets {2,4,8,12,16}\{2,4,8,12,16\} is 3.50±0.373.50\pm0.37 points. A three-seed MiniBooNE study is mixed at small budgets but positive at 8 and 16 acquisitions, identifying a current boundary rather than supporting a universal claim. These results establish a reproducible mechanism-level case for direct Bellman risk regression and delimit the experiments still needed for state-of-the-art comparison.

Explore similar work

CardsList
  1. Non-Myopic Active Feature Acquisition via Pathwise Policy Gradients

    May 6, 2026Linus Aronsson, Morteza Haghir ChehreghaniActive Learning