Special Session: Approximation and Fault Resiliency of DNN Accelerators
This work addresses reliability assessment for DNN accelerators in safety-critical domains like autonomous driving, but it is incremental as it builds on existing approximation and fault injection methods.
The paper tackles the problem of evaluating the reliability of DNN accelerators for safety-critical applications by proposing the use of approximate arithmetic circuits to emulate errors without costly fault injection, resulting in an efficient GPU-based simulation framework for fast evaluation.
Deep Learning, and in particular, Deep Neural Network (DNN) is nowadays widely used in many scenarios, including safety-critical applications such as autonomous driving. In this context, besides energy efficiency and performance, reliability plays a crucial role since a system failure can jeopardize human life. As with any other device, the reliability of hardware architectures running DNNs has to be evaluated, usually through costly fault injection campaigns. This paper explores the approximation and fault resiliency of DNN accelerators. We propose to use approximate (AxC) arithmetic circuits to agilely emulate errors in hardware without performing fault injection on the DNN. To allow fast evaluation of AxC DNN, we developed an efficient GPU-based simulation framework. Further, we propose a fine-grain analysis of fault resiliency by examining fault propagation and masking in networks