Invertible Neural Networks for Trustworthy AI
Malgorzata Schwab, Ashis Kumer Biswas · 2023
This study combines research in machine learning and system engineering practices to conceptualize a paradigm enhancing trustworthiness of machine learning inference process. We explore the topic of reversibility in deep neural networks and introduce their anomaly detection capabilities to build a framework of integrity verification checkpoints across the inference pipeline of a deployed model. We leverage previous findings and principles regarding several types of autoencoders, deep generative maximum-likelihood training and invertibility of neural networks to propose an improved network architecture for anomaly detection. A remarkable ability of an Invertible Neural Network (INN) to reconstruct data from its compressed representation and to solve inverse problems is then generalized and applied in the field of Trustworthy AI. To achieve integrity verification of an inference pipeline we place the INN-based Trusted Neural Network nodes around the mission critical parts of the system, achieving an end-to-end outcome verification. This work aspires to enhance robustness and reliability of applications employing artificial intelligence, which are playing increasingly noticeable role in highly consequential decision-making processes across many industries and problem domains.