ABSTRACT
Linear transformations are the dominating computation within many artificial intelligence (AI) applications. The natural multiply and accumulate feature of resistive crossbar arrays promise unprecedented processing capabilities to resistive dot-product engines (DPEs), which can accelerate approximate matrix-vector multiplication using analog in-memory computing. Unfortunately, the functional correctness of the accelerated AI applications may be compromised by various sources of errors. In this paper, we will outline the most pressing robustness challenges, the limitations of state-of-the-art solutions, and future opportunities for research.
Supplemental Material
- [n.d.]. Exascale Proxy Applications. https://proxyapps.exascaleproject.org.Google Scholar
- [n.d.]. Joint University Microelectronics Program (JUMP). https://www.darpa. mil/program/joint-university-microelectronics-program.Google Scholar
- Fabien Alibart, Ligang Gao, Brian D Hoskins, and Dmitri B Strukov. 2012. High Precision Tuning of State for Memristive Devices by Adaptable Variation-tolerant Algorithm. Nanotechnology, Vol. 23, 7 (2012), 075201.Google ScholarCross Ref
- Cornelia I Bargmann and William T Newsome. 2014. The Brain Research Through Advancing Innovative Neurotechnologies (BRAIN) Initiative and Neurology. JAMA neurology, Vol. 71, 6 (2014), 675--676.Google Scholar
- Karsten Beckmann, Josh Holt, Harika Manem, Joseph Van Nostrand, and Nathaniel C Cady. 2016. Nanoscale Hafnium Oxide RRAM Devices Exhibit Pulse Dependent Behavior and Multi-level Resistance Capability. Mrs Advances, Vol. 1, 49 (2016), 3355--3360.Google ScholarCross Ref
- Mahdi Nazm Bojnordi and Engin Ipek. 2016. Memristive Boltzmann Machine: A Hardware Accelerator for Combinatorial Optimization and Deep Learning. In HPCA'16. IEEE, 1--13.Google Scholar
- Ting Chang, Sung-Hyun Jo, and Wei Lu. 2011. Short-Term Memory to Long-Term Memory Transition in a Nanoscale Memristor. ACS nano, Vol. 5, 9 (2011), 7669--7676.Google Scholar
- Lerong Chen, Jiawen Li, Yiran Chen, Qiuping Deng, Jiyuan Shen, Xiaoyao Liang, and Li Jiang. 2017. Accelerator-friendly Neural-network Training: Learning Variations and Defects in RRAM Crossbar. In DATE'17. IEEE, 19--24.Google Scholar
- Ping Chi, Shuangchen Li, Cong Xu, Tao Zhang, Jishen Zhao, Yongpan Liu, Yu Wang, and Yuan Xie. 2016. PRIME: A Novel Processing-in-memory Architecture for Neural Network Computation in ReRAM-based Main Memory. ACM SIGARCH Computer Architecture News, Vol. 44, 3 (2016), 27--39.Google ScholarDigital Library
- Shinhyun Choi, Yuchao Yang, and Wei Lu. 2014. Random telegraph noise and resistance switching analysis of oxide based resistive memory. Nanoscale, Vol. 6, 1 (2014), 400--404.Google ScholarCross Ref
- Leon Chua. 1971. Memristor-The Missing Circuit Element. IEEE Transactions on circuit theory, Vol. 18, 5 (1971), 507--519.Google ScholarCross Ref
- R Degraeve, A Fantini, N Raghavan, L Goux, S Clima, B Govoreanu, A Belmonte, D Linten, and M Jurczak. 2015. Causes and consequences of the stochastic aspect of filamentary RRAM. Microelectronic Engineering, Vol. 147 (2015), 171--175.Google ScholarDigital Library
- Ben Feinberg, Uday Kumar Reddy Vengalam, Nathan Whitehair, Shibo Wang, and Engin Ipek. 2018. Enabling scientific computing on memristive accelerators. In ISCA'18. IEEE, 367--382.Google ScholarDigital Library
- Zhezhi He, Jie Lin, Rickard Ewetz, Jiann-Shiun Yuan, and Deliang Fan. 2019. Noise Injection Adaption: End-to-end ReRAM Crossbar Non-ideal Effect Adaption for Neural Network Mapping. In DAC'19. 1--6.Google ScholarDigital Library
- Miao Hu, Catherine E Graves, Can Li, Yunning Li, Ning Ge, Eric Montgomery, Noraica Davila, Hao Jiang, R Stanley Williams, J Joshua Yang, et almbox. 2018. Memristor-Based Analog Computation and Neural Network Classification with a Dot Product Engine. Advanced Materials, Vol. 30, 9 (2018), 1705914.Google ScholarCross Ref
- Miao Hu, Hai Li, Yiran Chen, Qing Wu, Garrett S Rose, and Richard W Linderman. 2014. Memristor Crossbar-Based Neuromorphic Computing System: A Case Study. IEEE TNNLS, Vol. 25, 10 (2014), 1864--1878.Google Scholar
- Miao Hu, John Paul Strachan, Zhiyong Li, Emmanuelle M Grafals, Noraica Davila, Catherine Graves, Sity Lam, Ning Ge, Jianhua Joshua Yang, and R Stanley Williams. 2016. Dot-Product Engine for Neuromorphic Computing: Programming 1T1M Crossbar to Accelerate Matrix-Vector Multiplication. In ACM/EDAC/IEEE DAC'16. IEEE, 1--6.Google ScholarDigital Library
- Yiming Huai. 2008. Spin-Transfer Torque MRAM (STT-MRAM): Challenges and Prospects. AAPPS bulletin, Vol. 18, 6 (2008), 33--40.Google Scholar
- Shubham Jain, Abhronil Sengupta, Kaushik Roy, and Anand Raghunathan. 2020. RxNN: A Framework for Evaluating Deep Neural Networks on Resistive Crossbars. IEEE TCAD (2020).Google Scholar
- Brian G Johnson and Charles H Dennison. 2004. Phase change memory. US Patent 6,791,102.Google Scholar
- Manuel Le Gallo, Abu Sebastian, Roland Mathis, Matteo Manica, Heiner Giefers, Tomas Tuma, Costas Bekas, Alessandro Curioni, and Evangelos Eleftheriou. 2018. Mixed-Precision In-Memory Computing. Nature Electronics, Vol. 1, 4 (2018), 246--253.Google ScholarCross Ref
- Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. 2015. Deep learning. Nature, Vol. 521, 7553 (2015), 436--444.Google Scholar
- HY Lee, PS Chen, TY Wu, YS Chen, CC Wang, PJ Tzeng, CH Lin, F Chen, CH Lien, and M-J Tsai. 2008. Low Power and High Speed Bipolar Switching with A Thin Reactive Ti Buffer Layer in Robust HfO$_2$ Based RRAM. In IEEE IEDM'08. IEEE, 1--4.Google Scholar
- Can Li, Yunning Li, Hao Jiang, Wenhao Song, Peng Lin, Zhongrui Wang, J Joshua Yang, Qiangfei Xia, Miao Hu, Eric Montgomery, et almbox. 2018. Large Memristor Crossbars for Analog Computing. In ISCAS'18. IEEE, 1--4.Google Scholar
- Beiye Liu, Miao Hu, Hai Li, Zhi-Hong Mao, Yiran Chen, Tingwen Huang, and Wei Zhang. 2013. Digital-Assisted Noise-Eliminating Training for Memristor Crossbar-based Analog Neuromorphic Computing Engine. In ACM/EDAC/IEEE DAC'13. IEEE, 1--6.Google Scholar
- Beiye Liu, Hai Li, Yiran Chen, Xin Li, Tingwen Huang, Qing Wu, and Mark Barnell. 2014. Reduction and IR-drop Compensations Techniques for Reliable Neuromorphic Computing Systems. In ICCAD'14. IEEE, 63--70.Google Scholar
- Beiye Liu, Hai Li, Yiran Chen, Xin Li, Qing Wu, and Tingwen Huang. 2015a. Vortex: Variation-aware Training for Memristor X-bar. In ACM/EDAC/IEEE DAC'15. 1--6.Google ScholarDigital Library
- Chenchen Liu, Miao Hu, John Paul Strachan, and Hai Li. 2017. Rescuing Memristor-based Neuromorphic Design with High Defects. In ACM/EDAC/IEEE DAC'17. IEEE, 1--6.Google Scholar
- Xiaoxiao Liu, Mengjie Mao, Beiye Liu, Hai Li, Yiran Chen, Boxun Li, Yu Wang, Hao Jiang, Mark Barnell, Qing Wu, et almbox. 2015b. RENO: A High-efficient Reconfigurable Neuromorphic Computing Accelerator Design. In DAC'15. 1--6.Google Scholar
- Yun Long, Xueyuan She, and Saibal Mukhopadhyay. 2019. Design of Reliable DNN Accelerator with Un-reliable ReRAM. In DATE'19. IEEE, 1769--1774.Google Scholar
- Stuart Parkin, Xin Jiang, Christian Kaiser, Alex Panchula, Kevin Roche, and Mahesh Samant. 2003. Magnetically Engineered Spintronic Sensors and Memory. Proc. IEEE, Vol. 91, 5 (2003), 661--680.Google ScholarCross Ref
- Ali Shafiee, Anirban Nag, Naveen Muralimanohar, Rajeev Balasubramonian, John Paul Strachan, Miao Hu, R Stanley Williams, and Vivek Srikumar. 2016. ISAAC: A Convolutional Neural Network Accelerator with In-Situ Analog Arithmetic in Crossbars. ACM SIGARCH Computer Architecture News, Vol. 44, 3 (2016), 14--26.Google ScholarDigital Library
- Linghao Song, Xuehai Qian, Hai Li, and Yiran Chen. 2017. PipeLayer: A Pipelined ReRAM-Based Accelerator for Deep Learning. In HPCA'17. IEEE, 541--552.Google Scholar
- Linghao Song, Youwei Zhuo, Xuehai Qian, Hai Li, and Yiran Chen. 2018. GraphR: Accelerating Graph Processing Using ReRAM. In HPCA'18. IEEE, 531--543.Google Scholar
- Dmitri B Strukov, Gregory S.Snider, Duncan R.Stewart, and R.Stanley Williams. 2009. The missing memristor found. Nature, Vol. 453, 12 (2009), 80--83.Google ScholarCross Ref
- Necati Uysal, Baogang Zhang, Sumit Kumar Jha, and Rickard Ewetz. 2020. DP-MAP: Towards Resistive Dot-Product Engines with Improved Precision. In IEEE/ACM ICCAD'20. IEEE, 1--9.Google ScholarDigital Library
- A van de Goor and Y Zorian. 1993. Effective March Algorithms for Testing Single-Order Addressed Memories. In 1993 European Conference on Design Automation with the European Event in ASIC Design. IEEE, 499--505.Google Scholar
- H-S Philip Wong, Heng-Yuan Lee, Shimeng Yu, Yu-Sheng Chen, Yi Wu, Pang-Shiu Chen, Byoungil Lee, Frederick T Chen, and Ming-Jinn Tsai. 2012. Metal--oxide RRAM. Proc. IEEE, Vol. 100, 6 (2012), 1951--1970.Google ScholarCross Ref
- H-S Philip Wong, Simone Raoux, SangBum Kim, Jiale Liang, John P Reifenberg, Bipin Rajendran, Mehdi Asheghi, and Kenneth E Goodson. 2010. Phase Change Memory. Proc. IEEE, Vol. 98, 12 (2010), 2201--2227.Google ScholarCross Ref
- Wm A Wulf and Sally A McKee. 1995. Hitting the Memory Wall: Implications of the Obvious. ACM SIGARCH computer architecture news, Vol. 23, 1 (1995), 20--24.Google ScholarDigital Library
- Lixue Xia, Peng Gu, Boxun Li, Tianqi Tang, Xiling Yin, Wenqin Huangfu, Shimeng Yu, Yu Cao, Yu Wang, and Huazhong Yang. 2016. Technological Exploration of RRAM Crossbar Array for Matrix-Vector Multiplication. JCST, Vol. 31, 1 (2016), 3--19.Google ScholarCross Ref
- Lixue Xia, Wenqin Huangfu, Tianqi Tang, Xiling Yin, Krishnendu Chakrabarty, Yuan Xie, Yu Wang, and Huazhong Yang. 2018. Stuck-at Fault Tolerance in RRAM Computing Systems. IEEE JETCAS, Vol. 8, 1 (2018), 102--115.Google Scholar
- L. Xia, Mengyun Liu, Xuefei Ning, K. Chakrabarty, and Yu Wang. 2017. Fault-Tolerant Training with On-Line Fault Detection for RRAM-Based Neural Computing Systems. In Proc. Design Automation Conference. 1--6.Google ScholarDigital Library
- Baogang Zhang and Rickard Ewetz. 2020. Towards Resilient Deployment of In-Memory Neural Networks with High Throughput. In DAC'21. 1--9.Google Scholar
- Baogang Zhang, Necati Uysal, and Rickard Ewetz. 2019 a. STAT: Mean and Variance Characterization for Robust Inference of DNNs on Memristor-based Platforms. In GLSVLSI. 339--342.Google Scholar
- Baogang Zhang, Necati Uysal, Deliang Fan, and Rickard Ewetz. 2019 b. Handling Stuck-at-Fault Defects Using Matrix Transformation for Robust Inference of DNNs. IEEE TCAD, Vol. 39, 10 (2019), 2448--2460.Google Scholar
- Baogang Zhang, Necati Uysal, Deliang Fan, and Rickard Ewetz. 2020. Representable Matrices: Enabling High Accuracy Analog Computation for Inference of DNNs using Memristors. In ASP-DAC'20. IEEE, 538--543.Google Scholar
- Fan Zhang and Miao Hu. 2020. Defects Mitigation in Resistive Crossbars for Analog Vector Matrix Multiplication. In ASP-DAC'20. IEEE, 187--192.Google Scholar
Index Terms
- Accelerating AI Applications using Analog In-Memory Computing: Challenges and Opportunities
Recommendations
Area-Efficient and Reliable Error Correcting Code Circuit Based on Hybrid CMOS/Memristor Circuit
AbstractResistive random-access memory (ReRAM) has several attractive features such as high storage density and high switching frequency with low power consumption. It is hence regarded as the most promising nonvolatile memory material. However, a ...
System-level impacts of persistent main memory using a search engine
Computer memory systems traditionally use distinct technologies for different hierarchy levels, typically volatile, high speed, high cost/byte solid state memory for caches and main memory (SRAM and DRAM), and non-volatile, low speed, low cost/byte ...
A novel tree-based representation for evolving analog circuits and its application to memristor-based pulse generation circuit
AbstractWhen applying evolutionary algorithms to circuit design automation, circuit representation is the first consideration. There have been several studies applying different circuit representations. However, they still have some problems, such as lack ...
Comments