當生醫晶片應用於生物體上時，會受到生物體自身雜訊的影響，使得每次量測到的訊號都不太一樣，因此無法準確的判定訊號是否為正常的訊號。為了克服這個困難，我們使用由Movellan於2002年所提出的擴散網路(Diffusion Network)演算法，作為研發辨識高維度時變函數的生醫訊號隨機晶片系統。擴散網路演算法是由數個隨機運算元所構成的類神經網路，運算元的運算是由時間函數的隨機微分方程式定義。在運算過程中會在方程式中加入雜訊項目，使得雜訊成為運算的一部分，因此擴散網路演算法可以提供即時的訊號變化以及對雜訊能保持一定的穩定性。
擴散網路演算法起初在本實驗室是藉由Matlab數學軟體來驗證其可行性，經過前幾位學長的努力，現行階段是應用VLSI技術來實現擴散網路晶片系統。在這個晶片中使用到類比乘法器、可變電阻、雜訊產生器、Sigmoid電路…等等來實現擴散網路晶片的訊號重建。由於前一版晶片並沒有成功重建出學習訊號，為找出原因，我們透過FIB(聚焦離子束)將晶片中的電路各別分開，量測各子電路的特性，找出不適用的電路並探討原因，再設計功能相同的新電路取代該電路。並且考慮整體電路的相容性，所以對整體電路進行模擬和討論是否會有互相干擾的問題。最後透過VLSI技術把電路實現成晶片系統，藉由晶片的量測來探討真實電路和模擬之間的誤差，以及解決的辦法。希望修正後的電路可以順利重建訊號，而未來則是把擴散網路作為模組化晶片，建立晶片之間溝通的橋樑，進而可以辨識更複雜的訊號。

Bibliography
[1] J. R. Movellan, P. Mineiro, and R. J. Williams, "A Monte Carlo EM Approach for partially observable diffusion processes: Theory and applications to neural networks," Neural Comput., vol. 14, no. 7 pp. 1507-1544, 2002.
[2] H. Chen and A. F. Murray, "A Continuous Restricted Boltzmann Machine with a Hardware-Amenable Learning Algorithm," presented at the Proceedings of the International Conference on Artificial Neural Networks, 2002.
[3] C. H. Chien, "A Stochastic System on a Chip Basing on Diffusion Network," Master's thesis, National Tsing Hua University, Taiwan, 2008.
[4] T. M. Kao, "Exploring the feasibility of training Diffusion Network with on-chip circuitry," Master's thesis, National Tsing Hua University, Taiwan, 2010.
[5] J. R. Movellan, "A learning theorem for networks at detailed stochastic equilibrium," Neural Computation, vol. 10, pp. 1157-1178, July 1998.
[6] Y.-S. Hsu, "Biomedical Signal Recognition Using Diffusion Networks," Master's thesis, National Tsing Hua University, Taiwan, 2007.
[7] C. Chen-Han, et al., "Mapping the Diffusion Network into a stochastic system in Very Large Scale Integration," in Neural Networks (IJCNN), The 2010 International Joint Conference on, pp. 1-7, 2010.
[8] K. Tanno, et al., "Four-quadrant CMOS current-mode multiplier independent of device parameters," Circuits and Systems II: Analog and Digital Signal Processing, IEEE Transactions on, vol. 47, pp. 473-477, 2000.
[9] S. F. Al-Sarawi, "A novel topology for grounded-to-floating resistor conversion in CMOS technology," Microelectronics Journal, vol. 33, pp. 1059-1069, 2002.
[10] G. Cauwenberghs, "An analog VLSI recurrent neural network learning a continuous-time trajectory," Neural Networks, IEEE Transactions on, vol. 7, pp. 346-361, 1996.
[11] F. Diotalevi, et al., "Analog CMOS current mode neural primitives," in Circuits and Systems, 2000. Proceedings. ISCAS 2000 Geneva. The 2000 IEEE International Symposium on, pp. 717-720 vol.2, 2000.
[12] M. Gravati, et al., "A novel current-mode very low power analog CMOS four quadrant multiplier," in Solid-State Circuits Conference,ESSCIRC. Proceedings of the 31st European, pp. 495-498, 2005.
[13] M. Gravati and M. Valle, "Modelling mismatch effects in CMOS translinear loops and current mode multipliers," in Circuit Theory and Design, Proceedings of the European Conference , pp. III/373-III/376 vol. 3, 2005.
[14] E. Vittoz and J. Fellrath, "CMOS analog integrated circuits based on weak inversion operations," Solid-State Circuits, IEEE Journal of, vol. 12, pp. 224-231, 1977.
[15] J. K. Shih-Chill Liu, Giacomo Indiveri, Tobias Delbruck,and Rodney Douglas, Analog VLSI:Circuits and Principlies, 2002.
[16] R. F. L. Rahul Sarpeshkar, and Carver Mead, "A low-power wide-linear-range transconductance amplifier," Analog Integr.Circuits Signal Process, vol. 13, pp. 123-151, 1997.
[17] W. Keng Hoong and R. Sarpeshkar, "An Electronically Tunable Linear or Nonlinear MOS Resistor," Circuits and Systems I: Regular Papers, IEEE Transactions on, vol. 55, pp. 2573-2583, 2008.
[18] G. Cauwenberghs, "Delta-sigma cellular automata for analog VLSI random vector generation," Circuits and Systems II: Analog and Digital Signal Processing, IEEE Transactions on, vol. 46, pp. 240-250, 1999.
[19] P.-L. Chen, "The development of embedded, current-mode non-volatile analog memory," Master's thesis, National Tsing Hua University, Taiwan, 2010.
[20] Y.-D. W. S.-J. L. H. Chen, "A Log-Domain Implementation of the Diffusion Network in Very Large Scale Integration," presented at the Advances in Neural Information Processing Systems Vancouver, British Columbia, Canada, 2010.