• 沒有找到結果。

管線化的倒傳遞類神經微處理器 謝元章、陳木松、陳慶順

N/A
N/A
Protected

Academic year: 2022

Share "管線化的倒傳遞類神經微處理器 謝元章、陳木松、陳慶順"

Copied!
2
0
0

加載中.... (立即查看全文)

全文

(1)

管線化的倒傳遞類神經微處理器 謝元章、陳木松、陳慶順

E-mail: 9510652@mail.dyu.edu.tw

摘 要

本研究運用演算法狀態機(Algorithmic State Machine, ASM)與 Verilog硬體描述語言(Hardware Description Language, HDL),

發展 一個管線化(Pipeline) 的倒傳遞類神經(Backpropagation Neural Network)的似MIPS架構之32位元微處理器。研究中 以MATLAB軟 體模擬類神經網路之運算,並推導似MIPS組合語言與機器碼,整 合入似MIPS 架構之32 位元精簡指令集微 處理器中。經由 SynaptiCAD 模擬的結果與MATLAB軟體模擬的結果相互比對驗 證,Xilinx FPGA晶片軟體合成,最後並 完成台積電0.18微米製程的 超大型積體電路佈局設計。

關鍵詞 : 演算法狀態機、管線化、倒傳遞類神經、MIPS、Verilog 目錄

封面內頁 簽名頁 授權書...iii 中文摘要...iv 英文摘

要...v 誌謝...vi 目錄...vii 圖目

錄...ix 表目錄...xii 第一章 緒論 1.1 研究動機...1 1.2 目的及方法...2 第二章 文獻回顧 2.1 類神經網路...3 2.2 精簡指令集電

腦...7 2.3 演算法狀態機...9 2.4 Verilog 硬體描述語言...12 第三章 研究方 法 3.1 MIPS CPU...14 3.2 資料路徑管線化...18 3.3 倒傳遞類神經網

路...21 第四章 結果與討論 4.1 設計流程...33 4.2 管線化的似MIPS微處理

器...34 4.3 浮點運算單元、自然指數函數模組...43 4.4 倒傳遞類神經網路測試...49 4.5 類 神經晶片模擬測試、合成驗證與IC 佈局...54 第五章 結論與建議 結論與建議...58 參考文

獻...60 附錄...64 參考文獻

[1] Rumelart, D.E., Hinton, G.E. and Williams, R.J. “Learning internalrepresentation by error propagation ”, in Parallel Distributed Processing, Vol.1,318-326,1986.

[2] Sung Su Kim, Seul Jung, “Hardware implementation of a real time neural network controller with a DSP and an FPGA”,Robotics and Automation, Proceedings. IEEE International Conference on, Volume 5 Page:4639 – 4644, 2004.

[3] Arnold, M.G., T.A. Bailey J.R. Cowles J.J. GupalF.N, “Behavior to Structure: Using Verilog andIn-Circuit Emulation to Teach How An AlgorithmBecomes Hardware”, Engineer, IEEE, Verilog HDL Conference, P19-28. 1995.

[4] Arnold, M.G. “Verilog Digital ComputerDesign Algorithms into Hardware”, 2001.

[5] Mark Holland, “Harnessin FPGAs for Computer Architecture Education, ”A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science In Electrical Engineering, University of Washington, 2002.

[6] 蔡安朝, 陳慶順, 潘天賜,“實現一個運用似MIPS架構之步進馬 達控制系統晶片”,中華民國自動控制研討會,大葉大學,2004。

[7] 劉俊佑, 陳慶順, 洪榮聰,“發展運用似MIPS架構之微控制 器”,中華民國自動控制研討會,大葉大學,2004。

[8] 林右文, 陳慶順, 呂嘉弘,“以FPGA實現交通號誌控制器”,中 華民國自動控制研討會,大葉大學,2004。

[9] Golden, M., Mudge, T., “Comparison of two common pipeline structures” Computers and Digital Techniques, IEE Proceedings , Volume:

143 , Issue: 3 ,Pages:161 - 167,1996.

[10] 王進德、蕭大全,“類神經網路與模糊控制理論入門”,全華科 技圖書股份有限公司,2003。

[11] Kab Joo Lee, “Fault sensitivity analysis of a 32-bit RISC microprocessor” VLSI and CAD, ICVC '99. 6th International Conference, pp.

529 - 532, 1999.

[12] Mark Gordon Arnold, “Verilog Digital Computer Design: Algorithms into Hardware”, Prentice Hall PTR, pp. 7-63, 1999.

[13] Ilya Levin, and Mark Karpovsky, “Synthesis of ASM-based Self Checking Controllers”, IEEE, pp. 87-93, 2001.

[14] CIC reference, Cell-Based IC Design Concepts, 2003.

[15] 曾志光 鄭光廷,“計算機組織與設計 軟硬體介面”,碁.,第 三、五章,3-2~5-43,2000。

[16] Liu Zhenyu , Qi Jiayue “Implementation of precise exception in a 5-stage pipeline embedded processor”ASIC, Proceedings. 5th

(2)

International Conference on Volume: 1 , 21-24,Pages:447 - 451 Vol.1, 2003.

[17] Rigler, A. K., Irvine, J. M. and T. P. Vogl, “Rescaling of variables in backpropagation learning, ” Neural Networks, vol. 4, pp. 225 -229, 1991.

[18] Barnard, E., “Optimization for training neural nets, ” IEEE Trans. Neural Network, vol. 3, pp. 232 -240, Mar , 1992.

[19] Hagan, M.T. and Menhaj, M.B. “Training feedforward networks with the Marquardt algorithm, ” IEEE Trans. Neural Network, vol. 5, pp. 989 -993, Nov., 1994.

[20] “Open Floating Point Unit” http://www.opencores.org.

[21] 張耀祖 白煌朗 方偉平,“數值分析”,松崗電腦圖書資料股份 有限公司,第三章,1991。

[22]H.T. Bui, B. Khalaf and S. Tahar: “Table-Driven Floating-Point Exponential Function” Technical Report, Concordia University,Department of Electrical and Computer Engineering, October 1998.

[23]Ping Tak Peter Tang. “Table-Driven Implementation of the Exponential Function in IEEE Floating-Point Arithmetic.” ACM Transactions on Mathematical Software. vol.no.2.1989.

[24] 葉怡成,“應用類神經網路” 儒林圖書有限公司,台北,1989。

[25] 林柏宏,碩士論文,“類神經網路應用於雙相氣泡流相傳遞特 性之即時決定”大葉大學,2002。

[26] Asai, T.; Yokotsuka, H.; Fukai, T.“A MOS circuit for a nonmonotonic neural network with excellent retrieval capabilities” Volume 7, Issue 1, Page:182 -189,1996.

[27] 曾榮鴻 莊季高,碩士論文,“類神經網路應用於數字識別之 分析與比較”,國立海洋大學,1992。

[28] Christopher M. Bishop “Exact Calculation of the Hessian Matrixfor the Multi-layer Perceptron” Published in Neural Computation 4 No.

4494 -501 , 1992.

[29] Guimaraes, F.G.; Ramirez, J.A.;Magnetics,“A Pruning Method for Neural Networks and ItsApplication for Optimization in Electromagnetics” IEEE Transactions onVolume 40, Part Page:1160-1163 Digital Object Identifier, 2004.

參考文獻

相關文件

Wang, Solving pseudomonotone variational inequalities and pseudo- convex optimization problems using the projection neural network, IEEE Transactions on Neural Network,

In this paper, we build a new class of neural networks based on the smoothing method for NCP introduced by Haddou and Maheux [18] using some family F of smoothing functions.

Numerical experiments are done for a class of quasi-convex optimization problems where the function f (x) is a composition of a quadratic convex function from IR n to IR and

Chen, The semismooth-related properties of a merit function and a descent method for the nonlinear complementarity problem, Journal of Global Optimization, vol.. Soares, A new

Qi (2001), Solving nonlinear complementarity problems with neural networks: a reformulation method approach, Journal of Computational and Applied Mathematics, vol. Pedrycz,

The Hilbert space of an orbifold field theory [6] is decomposed into twisted sectors H g , that are labelled by the conjugacy classes [g] of the orbifold group, in our case

Table 3 Numerical results for Cadzow, FIHT, PGD, DRI and our proposed pMAP on the noisy signal recovery experiment, including iterations (Iter), CPU time in seconds (Time), root of

The Model-Driven Simulation (MDS) derives performance information based on the application model by analyzing the data flow, working set, cache utilization, work- load, degree