1 / 54

Future Hardware Realization of Self-Organizing Learning Array and Its Software Simulation

Future Hardware Realization of Self-Organizing Learning Array and Its Software Simulation. Adviser: Dr. Janusz Starzyk Student: Tsun-Ho Liu. November 8 th , 2002. Ohio University School of Electrical Engineering and Computer Science. Outline. Introduction

ailsa
Download Presentation

Future Hardware Realization of Self-Organizing Learning Array and Its Software Simulation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Future Hardware Realization of Self-Organizing Learning Array and Its Software Simulation Adviser: Dr. Janusz Starzyk Student: Tsun-Ho Liu November 8th, 2002 Ohio University School of Electrical Engineering and Computer Science

  2. Outline • Introduction • Overview of Biological Neural Network • Self-Organizing Learning Array Structure • Neuron Structure and Self-Organizing Principles • Final Classification • Data Preprocessing • Software Simulation Result • Conclusion and Future Work School of Electrical Engineering and Computer Science

  3. Introduction • Digital computers are good at: • Fast arithmetic calculation • Precise software execution • Digital computers bed at: • Interacting with data from environment • Adapting to different conditions School of Electrical Engineering and Computer Science

  4. Introduction (Cont’d) • Advantage of Artificial Neural Networks: • Software free • Robust classification and pattern recognition • Recommendation of an action • Massive parallelism School of Electrical Engineering and Computer Science

  5. Introduction (Cont’d) • Research Objective: • Less interconnection • Self-organizing • Local Learning • Nonspecific classification School of Electrical Engineering and Computer Science

  6. Overview of Biological Neural Network • What are biological neurons? • Receiving information • Integrating information • Transmitting information • No homogeneous organization • Different shapes School of Electrical Engineering and Computer Science

  7. Overview of Biological Neural Network (Cont’d) • Structure and function of a neuron (Fraser, 1998, September) School of Electrical Engineering and Computer Science

  8. Overview of Biological Neural Network (Cont’d) • Neurons Communicate Modified (Fraser, 1998, September) School of Electrical Engineering and Computer Science

  9. Overview of Biological Neural Network (Cont’d) • Categories of neurons: • Long-axon cells • Carrying information for long distance • Further away connections • Short-axon cells • Integrating and processing information • Local connections School of Electrical Engineering and Computer Science

  10. Self-Organizing Learning Array Structure • Three sub-components • Input • Process layer • Output School of Electrical Engineering and Computer Science

  11. Self-Organizing Learning Array Structure (Cont’d) • Structure School of Electrical Engineering and Computer Science

  12. Self-Organizing Learning Array Structure (Cont’d) • Feed forward neural network organization School of Electrical Engineering and Computer Science

  13. Self-Organizing Learning Array Structure (Cont’d) • Initial Wiring School of Electrical Engineering and Computer Science

  14. Neuron Structure and Self-Organizing Principles • Neuron’s input and output School of Electrical Engineering and Computer Science

  15. Neuron Structure and Self-Organizing Principles (Cont’d) • Neuron Input • System clock • Data input • Threshold control input (TCI) • Input information deficiency School of Electrical Engineering and Computer Science

  16. Neuron Structure and Self-Organizing Principles (Cont’d) • System clock School of Electrical Engineering and Computer Science

  17. Neuron Structure and Self-Organizing Principles (Cont’d) • Data input School of Electrical Engineering and Computer Science

  18. Neuron Structure and Self-Organizing Principles (Cont’d) • Threshold control input (TCI) School of Electrical Engineering and Computer Science

  19. Neuron Structure and Self-Organizing Principles (Cont’d) • Input information deficiency • Indication of how much the input space (corresponding to this selected TCI) has been learned • [0 , 1] • 1 is set initially at the first input layer • 0 indicates this neuron has solved the problem 100% School of Electrical Engineering and Computer Science

  20. Neuron Structure and Self-Organizing Principles (Cont’d) • Termination of learning • Input information deficiency <= the chosen information deficiency threshold (IDT) • Try to learn from other subspace • If none, stop learning. School of Electrical Engineering and Computer Science

  21. Neuron Structure and Self-Organizing Principles (Cont’d) • Neuron inside • Transformation functions • Linear and nonlinear • Single input or multiple inputs • Information index calculation School of Electrical Engineering and Computer Science

  22. Neuron Structure and Self-Organizing Principles (Cont’d) School of Electrical Engineering and Computer Science

  23. Neuron Structure and Self-Organizing Principles (Cont’d) School of Electrical Engineering and Computer Science

  24. Neuron Structure and Self-Organizing Principles (Cont’d) • After optimum information index is obtained • Save: • Selected data inputs (or single input) • Selected threshold control input (TCI) • Selected transformation function • Selected threshold value • Probabilities of correct classification for different classes (passed threshold and does not passed threshold) School of Electrical Engineering and Computer Science

  25. Neuron Structure and Self-Organizing Principles (Cont’d) • Neuron output • System output • Output clocks • Threshold control output (TCO) • Threshold control output–threshold (TCOT) • Threshold control output–threshold–inverted (TCOTI) • Output information deficiencies School of Electrical Engineering and Computer Science

  26. Neuron Structure and Self-Organizing Principles (Cont’d) • System output School of Electrical Engineering and Computer Science

  27. Neuron Structure and Self-Organizing Principles (Cont’d) • Output Clock School of Electrical Engineering and Computer Science

  28. Neuron Structure and Self-Organizing Principles (Cont’d) • Output information deficiency • of TCO = Input information deficiency • of TCOT = Input information deficiency * local information deficiency (pass threshold) • of TCOTI = Input information deficiency * local information deficiency (does not pass threshold) School of Electrical Engineering and Computer Science

  29. Final Classification • Voting flags • Select-output-passed-threshold (SOT) • Set if output information deficiency of TCOT <= the voting threshold • Select-output-passed-threshold-inverse (SOTI) • Set if output information deficiency of TCOTI <= the voting threshold School of Electrical Engineering and Computer Science

  30. Final Classification (Cont’d) Pass Selected Threshold? Testing Data Selected TCI YES NO SOT=1? SOTI=1? Selected Transformation Function YES YES VOTE! NO NO END School of Electrical Engineering and Computer Science

  31. Final Classification (Cont’d) School of Electrical Engineering and Computer Science

  32. Further discussionConfidence interval School of Electrical Engineering and Computer Science

  33. Neuron 1 Neuron 2 Neuron 3 Neuron 4 Neuron 5 Neuron 6 Number of Samples 1 205 10 5 200 300 Pc for Class 1 1.00 0.60 0.71 0.08 0.12 0.25 Low Limit for Class 1 0.2065 0.5317 0.4057 0.007 0.0820 0.2044 High Limit for Class 1 1.00 0.6646 0.8978 0.5180 0.1723 0.3020 Calculated Mean 0.7676 0.5991 0.6793 0.1835 0.1237 0.2516 Pc for Class 2 0.00 0.40 0.29 0.92 0.88 0.75 Low Limit for Class 2 0.00 0.3354 0.1022 0.4820 0.8277 0.6980 High Limit for Class 2 0.7935 0.4683 0.5943 0.9930 0.9180 0.7956 Calculated Mean 0.2324 0.4009 0.3207 0.8165 0.8763 0.7484 Further discussionConfidence interval School of Electrical Engineering and Computer Science

  34. Further discussionConfidence interval • Condition A = Only Pc • Condition B = Mean of classes probabilities School of Electrical Engineering and Computer Science

  35. Data Preprocessing • Missing data recovery • All features are independent • Some features are dependent • Symbolic values assignment • Number of numerical feature = 1 • Number of numerical features > 1 School of Electrical Engineering and Computer Science

  36. Original Weight Recovered Weight 0.2800 0.4818 1.5100 1.4995 1.2945 1.1037 0.6995 0.7079 0.6880 0.8776 1.1000 0.9907 0.5780 0.9907 0.9070 0.7645 0.9615 0.8210 1.2960 0.9341 Missing data –independent features (ICS, UCI, 1995, December) School of Electrical Engineering and Computer Science

  37. Missing data –independent features School of Electrical Engineering and Computer Science

  38. #1 #2 #3 #4 #5 #6 #7 #8 ? 5699.7 5708.9 5905.6 4636.7 4555.4 5865.5 5724.2 ? 1827.0 1727.0 1923.0 1101.0 1231.0 1983.0 1837.0 2749.0 2843.0 2213.0 2938.0 2302.0 2943.0 2837.0 2744.0 ? 3743.0 3853.0 3848.0 3434.0 3223.0 3748.0 3757.0 4360.0 4321.0 4996.0 4858.0 4324.0 4211.0 4983.0 4372.0 5700.0 5495.0 5323.0 5959.0 5483.0 5321.0 5848.0 5748.0 6210.0 6723.0 6232.0 6835.0 6859.0 6948.0 6382.0 6223.0 5682.8 5718.3 4788.4 5263.3 5891.1 6101.0 4912.0 56746 7410.0 7239.0 7221.0 7122.0 7473.0 7484.0 7223.0 7434.0 8318.0 8372.0 8843.0 8235.0 8243.0 8873.0 8332.0 8321.0 9328.0 9323.0 9122.0 9422.0 9483.0 9566.0 9222.0 9332.0 Missing data – dependent features • Row 1 = Row 2 * 1.03 + Row 4 * 1.02 School of Electrical Engineering and Computer Science

  39. #1 #2 #3 #4 #5 #6 #7 #8 5772.1 5699.7 5708.9 5905.6 4636.7 4555.4 5865.5 5724.2 1919.9 1827.0 1727.0 1923.0 1101.0 1231.0 1983.0 1837.0 2749.0 2843.0 2213.0 2938.0 2302.0 2943.0 2837.0 2744.0 3720.2 3743.0 3853.0 3848.0 3434.0 3223.0 3748.0 3757.0 4360.0 4321.0 4996.0 4858.0 4324.0 4211.0 4983.0 4372.0 5700.0 5495.0 5323.0 5959.0 5483.0 5321.0 5848.0 5748.0 6210.0 6723.0 6232.0 6835.0 6859.0 6948.0 6382.0 6223.0 5682.8 5718.3 4788.4 5263.3 5891.1 6101.0 4912.0 56746 7410.0 7239.0 7221.0 7122.0 7473.0 7484.0 7223.0 7434.0 8318.0 8372.0 8843.0 8235.0 8243.0 8873.0 8332.0 8321.0 9328.0 9323.0 9122.0 9422.0 9483.0 9566.0 9222.0 9332.0 Missing data – dependent features School of Electrical Engineering and Computer Science

  40. Symbolic value – numerical feature =1 1) 3) 2) 4) School of Electrical Engineering and Computer Science

  41. Symbolic value – numerical feature =1 • Symbolic value – numerical feature =1 Xs = [1.0 3.0 3.0 3.5 3.5 8.5 8.5 9.0 9.0 9.0]T School of Electrical Engineering and Computer Science

  42. Data Preprocessing (Cont’d) 1) 4) 2) 5) 3) School of Electrical Engineering and Computer Science

  43. Data Preprocessing (Cont’d) • Symbolic value – numerical feature > 1 Xs = [1.0 2.85 2.85 3.274 3.274 7.241 7.241 7.884 7.88 7.884]T School of Electrical Engineering and Computer Science

  44. Class 1 Class2 Class 3 Class 4 Class 5 Number of points 503 429 190 682 542 Software Simulation Result School of Electrical Engineering and Computer Science

  45. Software Simulation Result (Cont’d) School of Electrical Engineering and Computer Science

  46. Software Simulation Result (Cont’d) School of Electrical Engineering and Computer Science

  47. Software Simulation Result (Cont’d) School of Electrical Engineering and Computer Science

  48. Software Simulation Result (Cont’d) School of Electrical Engineering and Computer Science

  49. Data Classified as Class 1 Class 2 Class 3 Class 4 Class 5 Data from Class 1 0.8171 0 0.0040 0.1769 0.0020 Data from Class 2 0 0.9977 0.0023 0 0 Data from Class 3 0 0 0.9263 0 0.0737 Data from Class 4 0.1481 0.0103 0 0.8416 0 Data from Class 5 0 0 0 0 1 Software Simulation Result (Cont’d) School of Electrical Engineering and Computer Science

  50. Credit Card Approval School of Electrical Engineering and Computer Science

More Related