Professor Abhijit Chatterjee at ECE Fall 2023 Colloquium

Error Resilient AI Systems: Addressing Soft Errors, Security Threats and Manufacturing Variability Effects

In this talk, we study the problem of designing error-resilient neuromorphic systems where errors can stem from: (a) soft errors in computation of matrix-vector multiplications and neuron activations, (b) malicious trojan and adversarial security attacks and (c) effects of manufacturing process variations on analog crossbar arrays that can affect DNN accuracy. The core principle of error detection and correction relies on the use of embedded neuron checks using invariants derived from the statistics of nominal neuron activation patterns as well as algorithmic encoding techniques. Errors are corrected using probabilistic methods due to difficulties involved in exact error diagnosis. The effects of manufacturing process variations are handled through the use of compact tests from which DNN performance can be assessed using learning techniques. Experimental results on a variety of neuromorphic test systems: DNNs, spiking networks, transformers and reinforcement learning are presented.