Sepehr Tabrizchi
ABSTRACT
Recently, Intelligent IoT (IIoT), including various sensors, has gained significant attention due to its capability of sensing, deciding, and acting by leveraging artificial neural networks (ANN). Nevertheless, to achieve acceptable accuracy and high performance in visual systems, a power-delay-efficient architecture is required. In this paper, we propose an ultra-low-power processing in-sensor architecture, namely SenTer, realizing low-precision ternary multi-layer perceptron networks, which can operate in detection and classification modes. Moreover, SenTer supports two activation functions based on user needs and the desired accuracy-energy trade-off. SenTer is capable of performing all the required computations for the MLP's first layer in the analog domain and then submitting its results to a co-processor. Therefore, SenTer significantly reduces the overhead of analog buffers, data conversion, and transmission power consumption by using only one ADC. Additionally, our simulation results demonstrate acceptable accuracy on various datasets compared to the full precision models.
- [n. d.]. CBCL FACE DATABASE. http://www.ai.mit.edu/projects/cbcl.old/software-datasets/FaceData2.html. Accessed: 2022-09-30.Google Scholar
- Minhaz Abedin, Arman Roohi, Maximilian Liehr, Nathaniel Cady, and Shaahin Angizi. 2022. MR-PIPA: An Integrated Multilevel RRAM (HfO x)-Based Processing- In-Pixel Accelerator. IEEE Journal on Exploratory Solid-State Computational Devices and Circuits 8, 2 (2022), 59--67.Google ScholarCross Ref
- Shaahin Angizi, Sepehr Tabrizchi, and Arman Roohi. 2022. Pisa: A binaryweight processing-in-sensor accelerator for edge image processing. arXiv preprint arXiv:2202.09035 (2022).Google Scholar
- Stephen J Carey, Alexey Lopich, David RW Barr, Bin Wang, and Piotr Dudek. 2013. A 100,000 fps vision sensor with embedded 535GOPS/W 256× 256 SIMD processor array. In 2013 symposium on VLSI circuits. IEEE, C182-C183.Google Scholar
- Amir Gholami, Sehoon Kim, Zhen Dong, Zhewei Yao, Michael W Mahoney, and Kurt Keutzer. 2021. A survey of quantization methods for efficient neural network inference. arXiv preprint arXiv:2103.13630 (2021).Google Scholar
- Tzu-Hsiang Hsu, Yi-Ren Chen, Ren-Shuo Liu, Chung-Chuan Lo, Kea-Tiong Tang, Meng-Fan Chang, and Chih-Cheng Hsieh. 2020. A 0.5-V real-time computational CMOS image sensor with programmable kernel for feature extraction. IEEE Journal of Solid-State Circuits 56, 5 (2020), 1588--1596.Google ScholarCross Ref
- Woo-Tae Kim, Hyunkeun Lee, Jung-Gyun Kim, and Byung-Geun Lee. 2020. An on-chip binary-weight convolution CMOS image sensor for neural networks. IEEE Transactions on Industrial Electronics 68, 8 (2020), 7567--7576.Google ScholarCross Ref
- Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. 1998. Gradientbased learning applied to document recognition. Proc. IEEE 86, 11 (1998), 2278--2324.Google ScholarCross Ref
- Qin Li, Changlu Liu, Peiyan Dong, Yanming Zhang, Tong Li, Sheng Lin, Minda Yang, Fei Qiao, YanzhiWang, Li Luo, et al. 2021. NS-FDN: Near-sensor processing architecture of feature-configurable distributed network for beyond-real-time always-on keyword spotting. IEEE Transactions on Circuits and Systems I: Regular Papers 68, 5 (2021), 1892--1905.Google ScholarCross Ref
- Robert LiKamWa, Yunhui Hou, Julian Gao, Mia Polansky, and Lin Zhong. 2016. Redeye: analog convnet image sensor architecture for continuous mobile vision. ACM SIGARCH Computer Architecture News 44, 3 (2016), 255--266.Google ScholarDigital Library
- Herbert Robbins and Sutton Monro. 1951. A stochastic approximation method. The annals of mathematical statistics (1951), 400--407.Google Scholar
- Sepehr Tabrizchi et al. 2023. AppCiP: Energy-Efficient Approximate Convolution-in- Pixel Scheme for Neural Network Acceleration. IEEE JETCAS 13, 1 (2023), 225--236.Google Scholar
- Sepehr Tabrizchi, Shaahin Angizi, and Arman Roohi. 2022. TizBin: A Low-Power Image Sensor with Event and Object Detection Using Efficient Processing-in-Pixel Schemes. In 2022 IEEE 40th International Conference on Computer Design (ICCD). IEEE, 770--777.Google Scholar
- Han Xiao, Kashif Rasul, and Roland Vollgraf. 2017. Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms. arXiv preprint arXiv:1708.07747 (2017).Google Scholar
- Han Xu, Ziru Li, Ningchao Lin, Qi Wei, Fei Qiao, Xunzhao Yin, and Huazhong Yang. 2020. Macsen: A processing-in-sensor architecture integrating mac operations into image sensor for ultra-low-power bnn-based intelligent visual perception. IEEE Transactions on Circuits and Systems II: Express Briefs 68, 2 (2020), 627--631.Google ScholarCross Ref
- Tomohiro Yamazaki, Hironobu Katayama, Shuji Uehara, Atsushi Nose, Masatsugu Kobayashi, Sayaka Shida, Masaki Odahara, Kenichi Takamiya, Yasuaki Hisamatsu, Shizunori Matsumoto, et al. 2017. 4.9 A 1ms high-speed vision chip with 3Dstacked 140GOPS column-parallel PEs for spatio-temporal image processing. In 2017 IEEE International Solid-State Circuits Conference (ISSCC). IEEE, 82--83.Google ScholarCross Ref
Index Terms
- SenTer: A Reconfigurable Processing-in-Sensor Architecture Enabling Efficient Ternary MLP
Recommendations
A scalable pipelined architecture for real-time computation of MLP-BP neural networks
In this paper a novel architecture for implementing multi-layer perceptron (MLP) neural networks on field programmable gate arrays (FPGA) is presented. The architecture presents a new scalable design that allows variable degrees of parallelism in order ...
Modelling and FDI of Dynamic Discrete Time Systems Using a MLP with a New Sigmoidal Activation Function
In this paper we investigate the use of the multi-layer perceptron (MLP) for system modelling. A new sigmoidal activation function is introduced and the study is focused at the utilization of this function on a MLP that performs modelling of dynamic, ...
Noise Suppression by Artificial Neural Networks
AWICT 2017: Proceedings of the Second International Conference on Advanced Wireless Information, Data, and Communication TechnologiesIn this research paper, noise suppression ability of associative memories (e.g. Hopfield neural network) is briefly summarized. Motivated by this fact, noise suppression abilities of trained convolutional network are discussed. Introducing the concept ...
Comments