Goal-Oriented Influence-Maximizing Data Acquisition for Learning and Optimization

TL;DR

This paper introduces GOIMDA, a novel active data acquisition method that maximizes influence on specific goals without relying on explicit posterior inference, improving efficiency in learning and optimization tasks.

Contribution

GOIMDA is a new influence-maximizing active acquisition algorithm that is uncertainty-aware without Bayesian posterior, applicable to various learning and optimization problems.

Findings

Achieves target performance with fewer samples or evaluations.

Outperforms uncertainty-based active learning and Bayesian optimization baselines.

Works effectively across image, text classification, and optimization tasks.

Abstract

Active data acquisition is central to many learning and optimization tasks in deep neural networks, yet remains challenging because most approaches rely on predictive uncertainty estimates that are difficult to obtain reliably. To this end, we propose Goal-Oriented Influence- Maximizing Data Acquisition (GOIMDA), an active acquisition algorithm that avoids explicit posterior inference while remaining uncertainty-aware through inverse curvature. GOIMDA selects inputs by maximizing their expected influence on a user-specified goal functional, such as test loss, predictive entropy, or the value of an optimizer-recommended design. Leveraging first-order influence functions, we derive a tractable acquisition rule that combines the goal gradient, training-loss curvature, and candidate sensitivity to model parameters. We show theoretically that, for generalized linear models, GOIMDA approximates predictive-entropy minimization up to a correction term accounting for goal alignment and prediction bias, thereby, yielding uncertainty-aware behavior without maintaining a Bayesian posterior. Empirically, across learning tasks (including image and text classification) and optimization tasks (including noisy global optimization benchmarks and neural-network hyperparameter tuning), GOIMDA consistently reaches target performance with substantially fewer labeled samples or function evaluations than uncertainty-based active learning and Gaussian-process Bayesian optimization baselines.