Robust Computing for Machine Learning-Based Systems

Abstract

The drive for automation and constant monitoring has led to rapid development in the field of Machine Learning (ML). The high accuracy offered by the state-of-the-art ML algorithms like Deep Neural Networks (DNNs) has paved the way for these algorithms to being used even in the emerging safety-critical applications, e.g., autonomous driving and smart healthcare. However, these applications require assurance about the functionality of the underlying systems/algorithms. Therefore, the robustness of these ML algorithms to different reliability and security threats has to be thoroughly studied and mechanisms/methodologies have to be designed which result in increased inherent resilience of these ML algorithms. Since traditional reliability measures like spatial and temporal redundancy are costly, they may not be feasible for DNN-based ML systems which are already super computer and memory intensive. Hence, new robustness methods for ML systems are required. Towards this, in this chapter, we present our analyses illustrating the impact of different reliability and security vulnerabilities on the accuracy of DNNs. We also discuss techniques that can be employed to design ML algorithms such that they are inherently resilient to reliability and security threats. Towards the end, the chapter provides open research challenges and further research opportunities.

RescueSNN: enabling reliable executions on spiking neural network accelerators under permanent faults

“…To maximize the performance and energy efficiency of SNN processing, specialized SNN accelerators/chips are employed (Painkras et al, 2013 ; Akopyan et al, 2015 ; Davies et al, 2018 ; Frenkel et al, 2019 ). However, these SNN chips may suffer from permanent faults, which can occur during: (1) chip fabrication process due to manufacturing defects, as fabricating an SNN chip with millions-to-billions of nano-scale transistors with 100% correct functionality is difficult, and even worsen due to the aggressive technology scaling (Hanif et al, 2018 , 2021 ; Zhang et al, 2018 ); and (2) run time operation due to device/transistor wear out and damages, that are caused by Hot Carrier Injection (HCI), Bias Temperature Instability (BTI), electromigration, or Time Dependent Dielectric Breakdown (TDDB) (Radetzki et al, 2013 ; Werner et al, 2016 ; Hanif et al, 2018 , 2021 ; Baloch et al, 2019 ; Mercier et al, 2020 ).…”