隨著生物電子領域的快速發展，生物電子系統結合植入式元件（implantable device）在生醫方面的應用展現了相當的發展潛力。而在這種充滿雜訊的生醫環境下，機率型類神經演算法（stochastic neural computation）被認為能增加系統對計算錯誤的容忍度，因而被視為是建立嵌入式系統（embedded-system）的一個可靠方式。在這之中，配合「對比差異最小化」（Minimizing Contrastive Divergence）調變電路的「連續值侷限型波茲曼模型」（Continuous Restricted Boltzmann Machine）因為具有這種特色因而被認為具有被實踐於超大型積體電路上的潛力。然而，量測結果卻顯示這個調變電路由於缺乏足夠的精準度而無法成功運作。
因此，此篇論文探討在積體電路上實現的「對比差異最小化」運算是否能夠達到連續值侷限型波茲曼模型所要求的精準度。這同時也代表著將此模型實踐於超大型積體電路時直接做晶片上調變(On-chip Training)的可行性。
此研究將先探討CRBM系統在連續值侷限型波茲曼模型在處理實際生醫訊號時所能容忍的誤差，以及利用修正調變演算法以提高誤差容忍度的可能性。根據模擬歸納的所要求的精準度，此論文設計了幾種準確的調變電路，並模擬其誤差大小。將調變電路本身的誤差與連續值侷限型波茲曼模型能容忍的誤差比較，本論文探討對CRBM系統做晶片上調變的可行性，進階的討論指出除了改善製程差異所造成的誤差，必須再加上其他的方法才能夠繼續降低調變電路的誤差，使連續值侷限型波茲曼模型能夠被實踐於晶片系統上進而被更廣泛的利用於生醫應用。

Reference

[B1] M.Sun, M.Mickle, L.Wei, Q.Liu, and R.Sclabassi, "Data communication between brain implants and computer," IEEE Trans.Neural Syst, vol. 11, no. 2, pp. 189-192, 2003.
[B2] M.A.L.Nicolelis, "Actions from thoughts," Nature, vol. 409 pp. 403-407, 2001.
[B3] E.A.Johannessen, L.Wang, S.W.Reid, D.R.S.Cumming, L.Cui, T.B.Tang, M.Ahmadian, A.Astaras, P.Yam, A.F.Murray, B.W.Flynn, S.P.Beaumont, and J.M.Cooper, "Implementation of distributed sensors in a microsystems format," IEEE Trans.Biomed.Eng., vol. 51, no. 3, pp. 525-535, 2004.
[B4] K.Najafi and K.D.Wise, "An implantable multielectrode array with on-chip signal processing," IEEE J.Solid-State Circuits, vol. 21, no. 6, pp. 1035-1044, 1986.
[B5] T.B.Tang, M.Ahmadian, A.F.Murray, A.Astaras, L.Wang, J.M.Cooper, S.P.Beaumont, B.W.Flynn, and D.R.S.Cumming, "Toward a miniature wireless integrated multisensor microsystem for industrial and biomedical applications," IEEE Sens.J.Special Issue on Integr.Multisensor Syst.Signal Process., vol. 2, no. 6, pp. 628-635, 2002.
[B6] Chen, H. and Murray, A. F., "Continuous restricted Boltzmann machine with an implementable training algorithm," Iee Proceedings-Vision Image and Signal Processing, vol. 150, no. 3, pp. 153-158, 2003.
[B7] Hinton, G. E., "Training products of experts by minimizing contrastive divergence," Neural Computation, vol. 14, no. 8, pp. 1771-1800, 2002.
[B8] Chen H., Fleury P.C.D., and Murray A.F., "Continuous-Valued Probabilistic Behavior in a VLSI Generative Model," IEEE Transactions on Neural Networks : Accepted for future publication, no. 99, pp. 1-16, 2006.
[B9] Fleury, P., Chen, H., and Murray, A. F. On-Chip Contrastive Divergence Learning in analogue VLSI. 3, 1723-1728. 2005. Budapest, 2004IEEE,Proceedings of the International Joint Conference on Neural Networks . 7-25-0040.
Ref Type: Conference Proceeding
[B10] P.Smolensky, "Parallel Distributed Processing: Explorations in the Microstructure of Cognition," Information processing in dynamical systems: Foundations of harmony theory 1986, pp. 195-281.
[B11] H.Chen, P.Fleury, and A.F.Murray, "Minimising Contrastive Divergence in Noisy Mixed-mode VLSI Neurons," Advances in Neural Information Processing Systems 2004.
[B12] H.Chible. Analog circuit for synapse neural networks vlsi implementation. 2, 1004-1007. 2000. 7th IEEE Int. Conf. on Electronics, Circuits and Systems (ICECS 2000).
Ref Type: Conference Proceeding
[B13] R.Geiger, P.Allen, and N.Strader, VLSI Design Techniques for Analog and Digital Circuits, 1 ed. New York: McGraw-Hill Publishing Company, 1990.
[B14] A.F.Murray, "Novelty detection using products of simple experts: a potential architecture embedded system," Neural Netw., vol. 14 pp. 1257-1264, 2001.
[B15] TARASSENKO.L. and CLIFORD.G., "Detection of ectopic beats in the electrocardiogram using an autoassociative neural network," Neural Process Lett., vol. 14 pp. 15-25, 2001.
[B16] Wegmann, G. and Vittoz, E. A., "Analysis and improvements of accurate dynamic current mirrors," Solid-State Circuits, IEEE, vol. 25, no. 3, pp. 699-706, June1990.
[B17] Wegmann, G. and Vittoz, E. A., "Basic principles of accurate dynamic current mirrors," Circuits, Devices and Systems, IEE, vol. 137, no. 2, pp. 95-100, Apr.1990.
[B18] Wegmann, G. and Vittoz, E. A., "Very accurate dynamic current mirrors," Electronics Letters, vol. 25, no. 10, pp. 644-646, May1989.
[B19] Koichi Tanno, Okihiko Ishizuka, and Zheng Tang, "Four-quardrant CMOS current-mode multiplier independent of device parameter," Circuits and Systems II: Analog and Digital Signal Processing, IEEE, vol. 47, no. 5, pp. 473-477, 2000.
[B20] Takashi Morie, "Analog VLSI implementation of self-learning neural networks," in Gert Cauwenberghs and Magdy A.Bayoumi (eds.) Learning on Silicon::adaptive VLSI neural systems 1999, pp. 213-242.