Slumbering to Precision: Enhancing Artificial Neural Network Calibration Through Sleep-like Processes

TL;DR

This paper introduces Sleep Replay Consolidation (SRC), a biologically inspired, post-training calibration method for neural networks that improves confidence estimates by mimicking sleep-like internal replay, complementing existing techniques.

Contribution

The paper proposes a novel sleep-inspired calibration approach, SRC, which enhances neural network confidence estimates without retraining, advancing trustworthiness in AI models.

Findings

SRC improves calibration metrics like Brier score and entropy.

Combining SRC with temperature scaling yields superior calibration results.

SRC is competitive with standard calibration methods across multiple architectures.

Abstract

Artificial neural networks are often overconfident, undermining trust because their predicted probabilities do not match actual accuracy. Inspired by biological sleep and the role of spontaneous replay in memory and learning, we introduce Sleep Replay Consolidation (SRC), a novel calibration approach. SRC is a post-training, sleep-like phase that selectively replays internal representations to update network weights and improve calibration without supervised retraining. Across multiple experiments, SRC is competitive with and complementary to standard approaches such as temperature scaling. Combining SRC with temperature scaling achieves the best Brier score and entropy trade-offs for AlexNet and VGG19. These results show that SRC provides a fundamentally novel approach to improving neural network calibration. SRC-based calibration offers a practical path toward more trustworthy confidence estimates and narrows the gap between human-like uncertainty handling and modern deep networks.