1 / 14

ECE 488

ECE 488. Computer Engineering Design I Fall 2005 Hau Ngo Ming Zhang. Agenda. Administrative Discussion Syllabus Schedule etc. Project Discussion Milestone 1 (due today) Grouping. Administrative Discussion. http://www.lions.odu.edu/~vasari/ece488/index.html. Project Discussion.

Télécharger la présentation

ECE 488

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. ECE 488 Computer Engineering Design I Fall 2005 Hau Ngo Ming Zhang

  2. Agenda • Administrative Discussion • Syllabus • Schedule etc. • Project Discussion • Milestone 1 (due today) • Grouping

  3. Administrative Discussion http://www.lions.odu.edu/~vasari/ece488/index.html

  4. Project Discussion • Brain Characteristics • Approximately 1011 neurons and 1015 interconnections • A neuron receives, processes and transmits electrochemical signals dendrites axon synapses axons neuron . . . . . .

  5. wj0 wj1 wj(N-1) x0 x1  Netj yj f (.) xN-1 Project Discussion • Artificial Neural Network • Biologically inspired networks • Massively parallel architecture • Dense arrangements of interconnections and simple computing elements  f(.)

  6. Project Discussion • Adaptive Resonance Theory (ART) based Self-Organizing Neural Network • Capable of rapid stable learning of recognition categories • Example: • Categorize people based on: • Age • Gender • Height • Hair color • Eye color • Ethnicity, etc.

  7. Project Discussion • The learning and categorization algorithm of Binary-ART Network • For each input and a category j, calculate choice function Tj • Chosen category is indexed by J so that • For the chosen category, calculate the match function ρ’ • If chosen category meet vigilance criterion (ρ’≥ ρ), learning the new input by updating weight WJ • Else, select the next category calculate the match function ρ’ • If no match is found, new category is created

  8. Project Discussion Calculation ρ = 0.4 Input Since this is the first input, the input form the first category/group Weight if this category is I = 00001111 Category Weight (1) W1= 00001111 W1 = I

  9. Project Discussion Calculation ρ = 0.4 Input Calculate choice function with respect to Category 1 I = 11110000 Category Weight Choose max. choice function (1) W1= 00001111 Calculate match function for W1 (2) W2= 11110000 ρ’< ρ create a new category/group

  10. Project Discussion Calculation ρ = 0.4 Input Calculate choice function with respect to Category 1, 2 I = 01111000 Category Weight Choose max. choice function (1) W1= 00001111 Calculate match function for W2 (2) W2= 11110000 (2) W2= 01110000 ρ’≥ ρ I joins category 2; update W2

  11. Project Discussion • Specifications • Binary-ART Network • 8-bit input data • Design and implement ONE neuron in EPLD (limitation in resource) • Complete network with microcontroller PC Microcontroller (M68HC11) EPLD

  12. Milestones 1.   Test and verify EPLD programming hardware. Wire up the EPLD, program it using the ByteBlaster and verify the function of the programmed EPLD.  (August 29, 2005) 2.  Design and simulate the neural network processor in high level language e.g. C/C++ (September 12, 2005) 3.   Design and simulate a neuron circuit. The simulation results should show the EPLD input the data, read the synaptic weights, compute the net value, perform threshloding and modify the synaptic weights. (September 26, 2005) 4.   Build and test the neuron hardware on the EPLD. Program, debug and verify the capabilities listed in milestone 3. (October 17 , 2005) 5.    Interface the EPLD directly with the M68HC11. The hardware design should include way to monitor the internal operations. (October 31 , 2005) 6.    Build and test the complete neural processor. Interface with the M68HC11. In addition, develop a program that runs on the M68HC11 that can input any desired inputs into the neural processor. (November 21 , 2005) 7. Final report and project demonstration. Final report should include a description of the design, design processes, design decisions and design trade-offs presented in a clear and cohesive fashion. (December 5, 2005)

  13. Milestone 1 • Compile and Simulate a simple circuit (e.g. an AND gate) • Program circuit onto Max7000s chip • Verify the functionality of the circuit • Steps to program Max7000s chip can be found in the handouts • Demo?

  14. Grouping • Form groups of 3 • Groups are responsible for partition tasks among group members • Define responsibility each member clearly and early • One person in each group signs for: • Altera’s Board • Byteblaster Cable • Power Supply

More Related