Enhancing AI System Resiliency: Formulation and Guarantee for LSTM Resilience Based on Control Theory

TL;DR

This paper introduces a control theory-based framework to quantify and guarantee LSTM network resilience in control systems, focusing on recovery time after anomalies, with practical bounds and validation for safety-critical AI.

Contribution

It develops a new resilience metric and refines $ ext{delta}$ISS theory for LSTM, providing a data-independent upper bound on recovery time for resilience-aware training.

Findings

Effective resilience estimation demonstrated on simple models

Resilience-aware training improves recovery performance

Provides a foundation for safety-critical AI assurance

Abstract

This paper proposes a novel theoretical framework for guaranteeing and evaluating the resilience of long short-term memory (LSTM) networks in control systems. We introduce "recovery time" as a new metric of resilience in order to quantify the time required for an LSTM to return to its normal state after anomalous inputs. By mathematically refining incremental input-to-state stability ($\delta$ISS) theory for LSTM, we derive a practical data-independent upper bound on recovery time. This upper bound gives us resilience-aware training. Experimental validation on simple models demonstrates the effectiveness of our resilience estimation and control methods, enhancing a foundation for rigorous quality assurance in safety-critical AI applications.