Foundations of Reliable Inference: Reliability-Efficiency Co-Design

TL;DR

This paper explores a unified framework for achieving reliable AI inference by balancing trustworthy uncertainty estimates with computational efficiency.

Contribution

It introduces a co-design approach that integrates reliability and efficiency in Bayesian learning for AI models.

Findings

Develops a theoretical framework for reliability-efficiency co-design.

Addresses the challenge of computational overhead in trustworthy inference.

Proposes methods to balance uncertainty quantification with efficiency.

Abstract

Reliable inference requires that artificial intelligence (AI) models provide trustworthy uncertainty estimates, not merely accurate predictions. Recent advances in Bayesian learning have made significant progress toward this goal, and growing concerns about computational overhead have jointly shifted the design criterion from reliability alone to the co-design of reliability and efficiency, i.e., reducing computational overhead while preserving trustworthy uncertainty quantification. This thesis develops a unified framework from two perspectives to address the central question: can we efficiently perform reliable inference?