Bayesian Inference via GeTe_x-OTS Based Stochastic Synapse for Uncertainty-Aware Medical Diagnostics

Abstract

Neuromorphic architectures leveraging stochastic device physics present a transformative approach for implementing probabilistic computing paradigms capable of intrinsic uncertainty quantification. In this work, we present a germanium-telluride-based ovonic threshold switch (GeTe_x-OTS) exhibiting inherent stochastic dynamics, integrated into a compact 1-selector-1-transistor-1-resistor (1S1T1R) synaptic unit. The OTS devices are demonstrated within an 8K-array, confirming the future scalability for neuromorphic systems. Experimental validation on the GeTe_x devices shows stable Gaussian-distributed threshold voltage fluctuations (σ = 100 mV), enabling precise control of synaptic activation probability through input pulse modulation. Utilizing this intrinsic stochasticity, we implement mini hardware realization of a Monte Carlo Dropconnect (MC-Dropconnect) neural network, directly demonstrating the feasibility of the system. Applied to COVID-19 diagnosis using chest X-ray images, our system achieves robust uncertainty quantification through predictive entropy, improving classification accuracy to 98.1%, compared to 96.0% with a deterministic baseline. This uncertainty-aware hardware design strategy provides a scalable pathway for implementing energy-efficient neuromorphic systems with native uncertainty estimation capabilities.

Keywords

Monte Carlo dropconnect / ovonic threshold switch / probabilistic computing / stochastic synapse / uncertainty-aware

References

[1] M. Prezioso, F. Merrikh-Bayat, B. D. Hoskins, G. C. Adam, K. K. Likharev, and D. B. Strukov, “Training and Operation of an Integrated Neuromorphic Network Based on Metal-Oxide Memristors,” Nature 521, no. 7550 (2015): 61–64.

[2] P. M. Sheridan, F. Cai, C. Du, W. Ma, Z. Zhang, and W. D. Lu, “Sparse Coding With Memristor Networks,” Nature Nanotechnology 12, no. 8 (2017): 784–789.

[3] P. Yao, H. Wu, B. Gao, et al., “Fully Hardware-Implemented Memristor Convolutional Neural Network,” Nature 577, no. 7792 (2020): 641–646.

[4] Y. Zhang, P. Qu, Y. Ji, et al., “A System Hierarchy for Brain-Inspired Computing,” Nature 586, no. 7829 (2020): 378–384.

[5] Z. Yang, T. Wang, Y. Lin, et al., “A Vision Chip With Complementary Pathways for Open-World Sensing,” Nature 629, no. 8014 (2024): 1027–1033.

[6] B. Dang, T. Zhang, X. Wu, K. Liu, R. Huang, and Y. Yang, “Reconfigurable In-Sensor Processing Based on a Multi-Phototransistor–One-Memristor Array,” Nature Electronics 7, no. 11 (2024): 991–1003.

[7] G. Wu, X. Zhang, G. Feng, et al., “Ferroelectric-Defined Reconfigurable Homojunctions for In-Memory Sensing and Computing,” Nature Materials 22, no. 12 (2023): 1499–1506.

[8] G. W. Burr, R. M. Shelby, S. Sidler, et al. “Experimental Demonstration and Tolerancing of a Large-Scale Neural Network (165 000 Synapses), Using Phase-Change Memory as the Synaptic Weight Element,” IEEE Transactions on Electron Devices 62, no. 11 (2015): 3498–3507.

[9] Z. Wang, S. Joshi, S. E. Savel'ev, et al., “Memristors With Diffusive Dynamics as Synaptic Emulators for Neuromorphic Computing,” Nature Materials 16 (2017): 101–108.

[10] S. Ambrogio, S. Balatti, V. Milo, et al. Novel RRAM-Enabled 1T1R Synapse Capable of Low-Power STDP via Burst-Mode Communication and Real-Time Unsupervised Machine Learning. In 2016 IEEE Symposium on VLSI Technology; IEEE: Honolulu, HI, USA, 2016; pp 1–2.

[11] J. Pei, L. Deng, S. Song, et al., “Towards Artificial General Intelligence With Hybrid Tianjic Chip Architecture,” Nature 572, no. 7767 (2019): 106–111.

[12] W. Zhang, P. Yao, B. Gao, et al., “Edge Learning Using a Fully Integrated Neuro-Inspired Memristor Chip,” Science 381, no. 6663 (2023): 1205–1211.

[13] S. Wei, P. Yao, X. Guo, et al., “A 28-nm Static-Power-Free Fully Parallel Rram-Based TD CIM Macro With 1982 TOPS/W/Bit for Edge Applications,” IEEE Solid-State Circuits Letters 8 (2025): 21–24.

[14] S. Yan, J. Yue, C. He, et al., “A 28-nm Floating-Point Computing-in-Memory Processor Using Intensive-CIM Sparse-Digital Architecture,” IEEE Journal of Solid-State Circuits 59, no. 8 (2024): 2630–2643.

[15] H. Park, Y. Li, and R. W. Tsien, “Influence of Synaptic Vesicle Position on Release Probability and Exocytotic Fusion Mode,” Science 335, no. 6074 (2012): 1362–1366.

[16] J. G. G. Borst, “The Low Synaptic Release Probability In Vivo,” Trends in Neurosciences 33, no. 6 (2010): 259–266.

[17] T. Branco and K. Staras, “The Probability of Neurotransmitter Release: Variability and Feedback Control at Single Synapses,” Nature Reviews Neuroscience 10, no. 5 (2009): 373–383.

[18] A. A. Faisal, L. P. J. Selen, and D. M. Wolpert, “Noise in the Nervous System,” Nature Reviews Neuroscience 9, no. 4 (2008): 292–303.

[19] Y. Gal and Z. Ghahramani, “Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning.” International Conference on Machine Learning (PMLR, 2016), 1050–1059.

[20] Y. Lin, Q. Zhang, B. Gao, et al., “Uncertainty Quantification via a Memristor Bayesian Deep Neural Network for Risk-Sensitive Reinforcement Learning,” Nature Machine Intelligence 5, no. 7 (2023): 714–723.

[21] S. Dutta, G. Detorakis, A. Khanna, B. Grisafe, E. Neftci, and S. Datta, “Neural Sampling Machine With Stochastic Synapse Allows Brain-Like Learning and Inference,” Nature Communications 13, no. 1 (2022): 2571.

[22] Z. Ghahramani, “Probabilistic Machine Learning and Artificial Intelligence,” Nature 521, no. 7553 (2015): 452–459.

[23] F. Aguirre, A. Sebastian, M. Le Gallo, et al., “Hardware Implementation of Memristor-Based Artificial Neural Networks,” Nature Communications 15, no. 1 (2024): 1974.

[24] S. Lee, J. Lee, S. Kim, et al. Selector With Low Ioff (< 100 pA), Fast Switching Speed (Τd = 7 ns), and High Thermal Stability (400°C/30 min) for X-Point Memory Applications. Symposium on VLSI Technology Digest of Technical Papers 2021.

[25] W. Zhang, R. Mazzarello, M. Wuttig, and E. Ma, “Designing Crystallization in Phase-Change Materials for Universal Memory and Neuro-Inspired Computing,” Nature Reviews Materials 4, no. 3 (2019): 150–168.

[26] S. Ban, J. Lee, Y. Seo, W. Lee, T. Kim, and H. Hwang, “Advances in Ovonic Threshold Switch Selector Technologies for Storage Class Memory: From Fundamentals to Development and Beyond,” Advanced Electronic Materials 11, no. 6 (2025): 2400665.

[27] J. Yoo, I. Karpov, S. Lee, J. Jung, H. S. Kim, and H. Hwang, “Threshold Voltage Drift in Te-Based Ovonic Threshold Switch Devices Under Various Operation Conditions,” IEEE Electron Device Letters 41, no. 1 (2020): 191–194.

[28] M. Yamaguchi, R. Degraeve, D. Garbin, et al. Understanding the Cycling-Dependent Threshold Voltage Instability in OTS Devices. In 2022 International Electron Devices Meeting (IEDM); IEEE: San Francisco, CA, USA, 2022; pp 5.1.1-5.1.4.

[29] Z. Chai, W. Zhang, R. Degraeve, et al. Evidence of Filamentary Switching and Relaxation Mechanisms in Ge_xSe_1−x OTS Selectors. In 2019 Symposium on VLSI Technology; IEEE: Kyoto, Japan, 2019; pp T238–T239.

[30] F. Hatem, J. F. Zhang, J. Marsland, et al. Endurance Improvement of More than Five Orders in Ge_xSe_1−x OTS Selectors by Using a Novel Refreshing Program Scheme. In 2019 IEEE International Electron Devices Meeting (IEDM); IEEE: San Francisco, CA, USA, 2019; pp 35.2.1-35.2.4.

[31] Z. Chai, P. Freitas, W. D. Zhang, et al., “Stochastic Computing Based on Volatile GeSe Ovonic Threshold Switching Selectors,” IEEE Electron Device Letters 41, no. 10 (2020): 1496–1499.

[32] Z. Chai, P. Freitas, J. Marsland, et al., “GeSe-Based Ovonic Threshold Switching Volatile True Random Number Generator,” IEEE Electron Device Letters 41, no. 2 (2020): 228–231.

[33] D. Matsubayashi, S. Clima, T. Ravsher, et al. OTS Physics-Based Screening for Environment-Friendly Selector Materials. In 2022 International Electron Devices Meeting (IEDM); IEEE: San Francisco, CA, USA, 2022; pp 8.6.1-8.6.4.

[34] Z. Zhao, S. Clima, D. Garbin, et al., “Chalcogenide Ovonic Threshold Switching Selector,” Nano-Micro Letters 16, no. 1 (2024): 81.

[35] M. Abdar, F. Pourpanah, S. Hussain, et al., “A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges,” Information Fusion 76 (2021): 243–297.

[36] E. Begoli, T. Bhattacharya, and D. Kusnezov, “The Need for Uncertainty Quantification in Machine-Assisted Medical Decision Making,” Nature Machine Intelligence 1, no. 1 (2019): 20–23.

[37] C. Blundell, J. Cornebise, K. Kavukcuoglu, and D. Wierstra, “Weight Uncertainty in Neural Network.” International Conference on Machine Learning (PMLR, 2015), 1613–1622.

[38] A. Filos, S. Farquhar, A. N. Gomez, et al. A Systematic Comparison of Bayesian Deep Learning Robustness in Diabetic Retinopathy Tasks. arXiv December 22, 2019.

[39] Y. Lin, B. Gao, J. Tang, Q. Zhang, H. Qian, and H. Wu, “Deep Bayesian Active Learning Using In-Memory Computing Hardware,” Nature Computational Science 5, no. 1 (2024): 27–36.

[40] Y. Lin, X. S. Hu, H. Qian, et al. Bayesian Neural Network Realization by Exploiting Inherent Stochastic Characteristics of Analog RRAM. In 2019 IEEE International Electron Devices Meeting (IEDM); IEEE: San Francisco, CA, USA, 2019; pp 14.6.1-14.6.4.

[41] D. Kim, W. Choi, J. Lee, and H. Hwang, “Inherent Stochasticity of Ovonic Threshold Switch for Neuronal Dropout of Edge-AI Hardware,” IEEE Electron Device Letters 44, no. 8 (2023): 1372–1375.

[42] J. Heo, S. Kim, S. Kim, and M. H. Kim, “Configurable Synaptic and Stochastic Neuronal Functions in ZnTe-Based Memristor for an RBM Neural Network,” Advanced Science 11 (2024): 2405768.

[43] M. Xiang, X. Xie, P. Savarese, X. Yuan, M. Maire, and Y. Li. Drop-Connect as a Fault-Tolerance Approach for RRAM-Based Deep Neural Network Accelerators. In 2024 IEEE 42nd VLSI Test Symposium(VTS); IEEE, Tempe, AZ, USA, 2024; pp 1–7.

[44] J. Zhou, X. Zhang, Y. Zhang, et al., “Hardware-Implemented Dropconnect Function for Energy-Efficient Neuromorphic Computing,” Advanced Functional Materials 35, no. 41 (2025): 2503452.

[45] M. I. Jordan, Z. Ghahramani, T. S. Jaakkola, and L. K. Saul, “An Introduction to Variational Methods for Graphical Models,” Machine Learning 37, no. 2 (1999): 183–233.

[46] S. Kullback and R. A. Leibler, “On Information and Sufficiency,” The Annals of Mathematical Statistics 22, no. 1 (1951): 79–86.

[47] Y. Gal, Uncertainty in Deep Learning. Doctoral Dissertation (University of Cambridge, 2016).

[48] M. D. Hoffman, D. M. Blei, C. Wang, and J. Paisley, “Stochastic Variational Inference,” Journal of Machine Learning Research: JMLR 14, no. 1 (2013): 1303–1347.

[49] Y. Fu, B. Dong, W.-C. Su, et al., “Enhancing LiAlO_x Synaptic Performance by Reducing the Schottky Barrier Height for Deep Neural Network Applications,” Nanoscale 12, no. 45 (2020): 22970–22977.

[50] Q. Zhang, H. Wu, P. Yao, et al., “Sign Backpropagation: An On-Chip Learning Algorithm for Analog RRAM Neuromorphic Computing Systems,” Neural Networks 108 (2018): 217–223.

[51] E. A. Casu, W. A. Vitale, N. Oliva, et al. Hybrid Phase-Change — Tunnel FET (PC-TFET) Switch With Subthreshold Swing < 10 mV/Decade and Sub-0.1 Body Factor: Digital and Analog Benchmarking. In 2016 IEEE International Electron Devices Meeting (IEDM); IEEE: San Francisco, CA, USA, 2016; pp 19.3.1-19.3.4.

[52] M. E. H. Chowdhury, T. Rahman, A. Khandakar, et al., “Can AI Help in Screening Viral and COVID-19 Pneumonia?,” IEEE Access 8 (2020): 132665–132676.

RIGHTS & PERMISSIONS

2026 The Author(s). SmartMat published by Tianjin University and John Wiley & Sons Australia, Ltd.