內存計算 (CIM) 設備因其卓越的性能和低功耗而受到越來越多的關注，特別是它們在神經網絡計算方面的新興功能。 然而，在 CIM 中集成模擬或混合信號等非數字信號可能會產生過程變化，從而導致計算不准確。 為了解決這個問題，我們引入了一種創新的矽後策略，提出了有效調整權重並最小化工藝變化的算法，並應用貝葉斯優化來微調模型參數。 這種優雅而靈活的解決方案可以自適應地解決各種噪聲環境，而無需模型重新訓練，並且其效率已通過我們在關鍵詞識別和圖像分類任務中的實驗得到了實證證實。 例如，在關鍵字識別 (KWS) 的二進制 CIM 實驗中，我們的方法將準確度從 58.18% 顯著提高到了令人印象深刻的 88.58%。 因此，我們相信我們提出的策略將極大地鼓勵採用模擬設備進行神經網絡計算。

Increasing attention is being drawn to Computing in Memory (CIM) devices due to their superior performance and low power consumption, especially with their burgeoning capabilities in neural network computations. However, integrating non-digital signals like analog or mixed ones within CIM could generate process variations, leading to computational inaccuracies. To tackle this issue, we've introduced an innovative post-silicon strategy proposing algorithms to effectively adjust weights and minimize process variations, along with the application of Bayesian optimization for the fine-tuning of model parameters. This elegant and flexible solution adaptively addresses various noise environments without requiring model retraining, and its efficiency has been empirically confirmed through our experiments in keyword spotting and image classification tasks. For example, in the binary CIM experiments for Keyword Spotting (KWS), our methodology significantly improved accuracy from 58.18% to an impressive 88.58%. Consequently, we are confident that our proposed strategy will substantially encourage the adoption of emulation devices for neural network computations.

Acknowledgements (Chinese) I
Abstract (Chinese) III
Abstract IV
Contents V
List of Figures VII
List of Tables VIII
List of Algorithms IX
1 Introduction 1
2 Backgorund 4
2.1 Computing-In-Memory (CIM) . . . . . . . . . . . . . . . . . . . . . 4
2.2 Variation impacts & modeling . . . . . . . . . . . . . . . . . . . . . 8
2.3 Related works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Proposed calibration methods 12
3.1 Variation characterization . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Calibration methods . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2.1 Binarize threshold adjustment . . . . . . . . . . . . . . . . . 14
3.2.2 Fusion method . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Enhancement method . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4 Experiments 21
4.1 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Experimental setting . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Experimental results . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.3.1 Effectiveness of calibration . . . . . . . . . . . . . . . . . . . 23
4.3.2 Comparison with noise injection training . . . . . . . . . . . 23
5 Conclusion 28
References 29

[1] Y.-C. Chiu, Z. Zhang, J.-J. Chen, X. Si, R. Liu, Y.-N. Tu, J.-W. Su, W.-H. Huang, J.-H. Wang, W.-C. Wei, J.-M. Hung, S.-S. Sheu, S.-H. Li, C.-I.Wu, R.-S. Liu, C.-C. Hsieh, K.-T. Tang, and M.-F. Chang,“A 4-kb 1-to-8-bitconfigurable 6t sram-based computation-in-memory unitmacro for cnn-based ai edge processors”, IEEE Journal of Solid-State Circuits, vol. 55, no. 10, pp. 2790–2801, Oct 2020.
[2] C.-S. Lin, F.-C. Tsai, J.-W. Su, S.-H. Li, T.-S. Chang, S.-S. Sheu, W.-C. Lo, S.-C. Chang, C.-I. Wu, and T.-H. Hou,“A 48 tops and 20943 tops/w 512kb computation-in-sram macro for highly reconfigurable ternary cnn acceleration”, in 2021 IEEE Asian Solid-State Circuits Conference (ASSCC), Nov 2021, pp. 1–3.
[3] C. Zhou, P. Kadambi, M. Mattina, and P. N. Whatmough,“Noisy
machines: Understanding noisy neural networks and enhancing robustness to analog hardware errors using distillation”, 2020.Online Available:
https://arxiv.org/abs/2001.04974
[4] J. Lorenz, E. Bar, A. Burenkov, P. Evanschitzky, A. Asenov, L. Wang, ¨ X. Wang, A. R. Brown, C. Millar, and D. Reid,“Simultaneous simulation of systematic and stochastic process variations”, in 2014 International Conference on Simulation of Semiconductor Processes and Devices (SISPAD), Sep. 2014, pp. 289–292.
[5] Q. Wang, Y. Park, and W. Lu,“Device variation effects on neural network inference accuracy in analog in-memory computing systems” , Advanced Intelligent Systems, pp. 2100199, 01 2022.
[6] Z.-H. Lee, F.-C. Tsai, and S.-C. Chang,“Robust binary neural network against noisy analog computation”, in 2022 Design, Automation Test in Europe Conference Exhibition (DATE), March 2022, pp. 484–489.
[7] Y.-W. Kang, C.-F. Wu, Y.-H. Chang, T.-W. Kuo, and S.-Y. Ho,“On minimizing analog variation errors to resolve the scalability issue of reram-based crossbar accelerators”, IEEE Transactions on ComputerAided Design of Integrated Circuits and Systems, vol. 39, no. 11, pp.3856–3867, Nov 2020.
[8] M. Qin and D. Vucinic, “Training recurrent neural networks against
noisy computations during inference”, 2018. Online Available:
https://arxiv.org/abs/1807.06555
[9] Y. Jiang, R. M. Zur, L. L. Pesce, and K. Drukker, “A study of the effect of noise injection on the training of artificial neural networks”, in 2009 International Joint Conference on Neural Networks, June 2009, pp. 1428–1432.
[10] V. Joshi, M. L. Gallo, S. Haefeli, I. Boybat, S. R. Nandakumar, C. Piveteau, M. Dazzi, B. Rajendran, A. Sebastian, and E. Eleftheriou, “Accurate deep neural network inference using computational phase change memory”, Nature Communications, vol. 11, no. 1, may 2020. Online Available: https://doi.org/10.1038%2Fs41467-020-16108-9
[11] R. Bruce, S. G. Sarwat, I. Boybat, C.-W. Cheng, W. Kim, S. Nandakumar, C. Mackin, T. Philip, Z. Liu, K. Brew, N. Gong, I. Ok, P. Adusumilli, K. Spoon, S. Ambrogio, B. Kersting, T. Bohnstingl, M. L. Gallo, A. Simon, N. Li, I. Saraf, J.-P. Han, L. Gignac, J. Papalia, T. Yamashita, N. Saulnier, G. W. Burr, H. Tsai, A. Sebastian, V. Narayanan, and M. BrightSky, “Mushroom-type phase change memory with projection liner: An array-level demonstration of conductance drift and noise mitigation”, in 2021 IEEE International Reliability Physics Symposium (IRPS), March 2021, pp. 1–6.
[12] S. R. Nandakumar, I. Boybat, V. Joshi, C. Piveteau, M. Le Gallo, B. Rajendran, A. Sebastian, and E. Eleftheriou, “Phase-change memory models for deep learning training and inference”, in 2019 26th IEEE International Conference on Electronics, Circuits and Systems (ICECS), Nov 2019, pp. 727–730.
[13] J Mockus, V Tiesis, and A Zilinskas, “The application of Bayesian methods for seeking the extremum”, in 1978, Towards Global Optimization, 2:117–129
[14] M. Rastegari, V. Ordonez, J. Redmon, and A. Farhadi, “Xnor-net: Imagenet classification using binary convolutional neural networks”, in Computer Vision – ECCV 2016, B. Leibe, J. Matas, N. Sebe, and M. Welling, Eds. Cham: Springer International Publishing, 2016, pp. 525–542.
[15] H. Kim, J.-H. Bae, S. Lim, S.-T. Lee, Y.-T. Seo, D. Kwon,
B.-G. Park, and J.-H. Lee, “Efficient precise weight tuning protocol considering variation of the synaptic devices and target accuracy”, Neurocomputing, vol. 378, pp. 189–196, 2020. Online Available:
https://www.sciencedirect.com/science/article/pii/S0925231219314900
[16] M. Klachko, M. R. Mahmoodi, and D. B. Strukov, “Improving
noise tolerance of mixed-signal neural networks”, 2019. Online Available: https://arxiv.org/abs/1904.01705
[17] C. Zhou, P. Kadambi, M. Mattina, and P. N. Whatmough, “Noisy
machines: Understanding noisy neural networks and enhancing robustness to analog hardware errors using distillation”, 2020. Online Available:
https://arxiv.org/abs/2001.04974
[18] P. Warden, “Speech commands: A dataset for limited-vocabulary speech recognition”, 2018. Online Available: https://arxiv.org/abs/1804.03209
[19] M. Courbariaux, Y. Bengio, and J.-P. David, “Binaryconnect: Training deep
neural networks with binary weights during propagations”, 2015. Online Available: https://arxiv.org/abs/1511.00363
[20] R. Tang and J. Lin, “Deep residual learning for small-footprint keyword spotting”, 2017. Online Available: https://arxiv.org/abs/1710.10361
[21] S. Nassif, “Modeling and forecasting of manufacturing variations”, in Proceedings of the ASP-DAC 2001. Asia and South Pacific Design Automation
Conference 2001 (Cat. No.01EX455), Feb 2001, pp. 145–149
[22] S. Yu, H. Jiang, S. Huang, X. Peng, and A. Lu, “Compute-in-memory chips
for deep learning: Recent trends and prospects”, IEEE Circuits and Systems
Magazine, vol. 21, no. 3, pp. 31–56, thirdquarter 2021.
[23] A. Krizhevsky, “Learning multiple layers of features from tiny images”, International Conference on Learning Representations (ICLR), 2015.
[24] C.-I. Chung, “Post-Silicon Calibration of CIM Deep Learning Model”, Master
thesis, 2022