Training Modern Deep Neural Networks for Memory-Fault Robustness
This addresses energy efficiency for DNN deployment in resource-limited hardware, but is incremental as it builds on existing robustness techniques.
The paper tackles the problem of deep neural networks being vulnerable to memory bit-cell faults caused by reduced supply voltage in energy-constrained systems, and proposes a regularizer that maintains accuracy while enabling energy savings, with experiments showing no reduction in accuracy.
Because deep neural networks (DNNs) rely on a large number of parameters and computations, their implementation in energy-constrained systems is challenging. In this paper, we investigate the solution of reducing the supply voltage of the memories used in the system, which results in bit-cell faults. We explore the robustness of state-of-the-art DNN architectures towards such defects and propose a regularizer meant to mitigate their effects on accuracy. Our experiments clearly demonstrate the interest of operating the system in a faulty regime to save energy without reducing accuracy.