Invariant Transformation and Resampling based Epistemic-Uncertainty Reduction

TL;DR

This paper introduces a resampling method using invariant transformations to reduce epistemic uncertainty in AI inference, improving accuracy by aggregating multiple transformed inputs.

Contribution

It proposes a novel resampling-based inference technique leveraging invariant transformations to mitigate epistemic uncertainty in trained AI models.

Findings

Resampling with invariant transformations improves inference accuracy.

The method effectively reduces epistemic uncertainty.

It offers a strategy to balance model size and performance.

Abstract

An artificial intelligence (AI) model can be viewed as a function that maps inputs to outputs in high-dimensional spaces. Once designed and well trained, the AI model is applied for inference. However, even optimized AI models can produce inference errors due to aleatoric and epistemic uncertainties. Interestingly, we observed that when inferring multiple samples based on invariant transformations of an input, inference errors can show partial independences due to epistemic uncertainty. Leveraging this insight, we propose a "resampling" based inferencing that applies to a trained AI model with multiple transformed versions of an input, and aggregates inference outputs to a more accurate result. This approach has the potential to improve inference accuracy and offers a strategy for balancing model size and performance.