A Low Power Mixed Signal Convolutional Neural Network for Deep Learning SoC
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Convolutional Neural Networks (CNNs) are getting fame due to their simpler design and higher performance. However, CNNs suffer from a large area and power consumption constraints. The multiply-and-accumulate (MAC) unit, which performs the convolution operation inside a CNN, consumes a significant amount of power consumption. In this study, we propose a mixed-signal approach for implementing analog MAC unit that can replace the digital MAC unit in CNNs. The Analog MAC unit architecture is constituted from binary weighted current steering digital-to-analog (DAC) circuit and capacitors. A digital parallel interface is designed to provide input image and filter values to the MAC unit. To realize a complete CNN model a low-power analog-to-digital (ADC) is then employed at the output to convert the final value back to a digital value. When a 3×3 convolution is performed, the analog MAC unit offers a 10.7% reduction in area and a 59.2% reduction in power consumption compared to its fully digital counterparts.
Abstract
Convolutional Neural Networks (CNNs) are getting fame due to their simpler design and higher performance. However, CNNs suffer from a large area and power consumption constraints. The multiply-and-accumulate (MAC) unit, which performs the convolution operation inside a CNN, consumes a significant amount of power consumption. In this study, we propose a mixed-signal approach for implementing analog MAC unit that can replace the digital MAC unit in CNNs. The Analog MAC unit architecture is constituted from binary weighted current steering digital-to-analog (DAC) circuit and capacitors. A digital parallel interface is designed to provide input image and filter values to the MAC unit. To realize a complete CNN model a low-power analog-to-digital (ADC) is then employed at the output to convert the final value back to a digital value. When a 3×3 convolution is performed, the analog MAC unit offers a 10.7% reduction in area and a 59.2% reduction in power consumption compared to its fully digital counterparts.
참고문헌 (17)
- A. Kankanhalli, "IoT and AI for Smart Government : A Research Agenda", Government Information Quarterly, 36(2), pp.304~309 (2019)
- R. A. Nawrocki, "A Mini Review of Neuromorphic Architectures and Implementations", IEEE Transactions on Electron Devices, 63(10), pp.3819~3829 (2016)
- J. C. Knight, "GPUs Outperform Current HPC and Neuromorphic Solutions in Terms of Speed and Energy When Simulating a Highly-Connected Cortical Model", Frontiers in Neuroscience, 12, pp.941~ (2018)
- "Neuromorphic Hardware: Trying to Put Brain into Chips" (2019)[URL]
- H. Kim, "Spiking Neural Network Using Synaptic Transistors and Neuron Circuits for Pattern Recognition with Noisy Images", IEEE Electron Device Letters, 39(4), pp.630~633 (2018)
- H. Tang, "Spike Counts Based Low Complexity SNN Architecture with Binary Synapse", IEEE Transactions on Biomedical Circuits and Systems, 13(6), pp.1664~1677 (2019)
- G. K. Chen, "A 4096-Neuron 1M-Synapse 3. 8-pJ", SOP Spiking Neural Network with On-Chip STDP Learning and Sparse Weights in 10-nm FinFET CMOS, 2019, pp.54(4), 992~1002
- M. S. Asghar, "A Low-Power Spiking Neural Network Chip Based on a Compact LIF Neuron and Binary Exponential Charge Injector Synapse Circuits", Sensors, 21(13), pp.4462~ (2021)[DOI]
- Y. A. Bachtiar, "Convolutional Neural Network and Maxpooling Architecture on Zynq SoC FPGA", International Symposium on Electronics and Smart Devices (ISESD), pp.1~5 (2019)
- S. Sabogal, "ReCoN : A Reconfigurable CNN Acceleration Framework for Hybrid Semantic Segmentation on Hybrid SoCs for Space Applications", IEEE Space Computing Conference (SCC), pp.41~52 (2019)
- Y. Halawani, "ReRAM-Based In-Memory Computing for Search Engine and Neural Network Applications", IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 9(2), pp.388~397 (2019)
- S.-S. Park, "CENNA : Cost-Effective Neural Network Accelerator", MDPI Electronics, 9(1) (2020)
- H. Kwon, "MAERI : Enabling Flexible Dataflow Mapping over DNN Accelerators via Reconfigurable Interconnects", SIGPLAN Not, 53(2), pp.461~475 (2018)
- F. N. Buhler, "A 16-channel Noise-Shaping Machine Learning Analog-Digital Interface", IEEE Symp. VLSI Circuits (2016)
- J. -O. Seo, "ARCHON : A 332. 7TOPS", W 5b Variation-Tolerant Analog CNN Processor Featuring Analog Neuronal Computation Unit and Analog Memory, 2022 (2022), 258~260
- M. S. Asghar, "A Digitally Controlled Analog kernel for Convolutional Neural Networks", 2021 18th International SoC Design Conference (ISOCC), pp.242~243 (2021)
- J. Zhu, "Analog Implementation of Reconfigurable Convolutional Neural Network Kernels", 2019 IEEE Asia Pacific Conference on Circuits and Systems (APCCAS), Bangkok, Thailand, 2019, pp.265~268 (2019)