Minerva: Enabling Low-Power, Highly-Accurate Deep Neural Network Accelerators

Citations Over TimeTop 1% of 2016 papers

Abstract

The continued success of Deep Neural Networks (DNNs) in classification tasks has sparked a trend of accelerating their execution with specialized hardware. While published designs easily give an order of magnitude improvement over general-purpose hardware, few look beyond an initial implementation. This paper presents Minerva, a highly automated co-design approach across the algorithm, architecture, and circuit levels to optimize DNN hardware accelerators. Compared to an established fixed-point accelerator baseline, we show that fine-grained, heterogeneous data type optimization reduces power by 1.5×, aggressive, in-line predication and pruning of small activity values further reduces power by 2.0×, and active hardware fault detection coupled with domain-aware error mitigation eliminates an additional 2.7× through lowering SRAM voltages. Across five datasets, these optimizations provide a collective average of 8.1× power reduction over an accelerator baseline without compromising DNN model accuracy. Minerva enables highly accurate, ultra-low power DNN accelerators (in the range of tens of milliwatts), making it feasible to deploy DNNs in power-constrained IoT and mobile devices.

Related Papers

• → Design and implementaton of SRAM macro unit(2017)8 cited
• → Variability-aware 7T SRAM circuit with low leakage high data stability SLEEP mode(2015)12 cited
• → SRAM on-chip monitoring methodology for high yield and energy efficient memory operation at near threshold voltage(2020)2 cited
• → Self-healing integrated circuits/systems in semiconductor nanometer technologies(2020)
• → Performance Modeling Sparse MTTKRP Using Optical Static Random Access Memory on FPGA(2022)