1 / 13

Minimize the implementation cost by rewiring

Speaker : Chia-Chun Lin Advisor : Chun-Yao Wang 2013/5/13. Minimize the implementation cost by rewiring. Outline. Introduction Previous work Minimize the implementation cost Choose the target wire The target wire is critical The target wire is not critical Proposed flow

Download Presentation

Minimize the implementation cost by rewiring

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. Speaker: Chia-Chun Lin Advisor: Chun-Yao Wang 2013/5/13 Minimize the implementation cost by rewiring

  2. Outline • Introduction • Previous work • Minimize the implementation cost • Choose the target wire • The target wire is critical • The target wire is not critical • Proposed flow • Experimental result

  3. Linear threshold gate • A linear threshold gate (LTG) is an n binary input and one binary output function defined as below: f = 1 if 0if • A network is composed of LTGs is called a threshold network. a b c 2 1 1 3 f

  4. An example of threshold network • An example of threshold network and its corresponding Boolean representation. • Def: An input in a single group LTG is critical if and only if this LTG will become useless after removing this input. a b c 2 1 1 3 f

  5. Previous work(1/2) f f 10 10 • Rectification Network Construction • Case 1: The target wire is not critical: • Case 2: The target wire is critical, and we rectify it at the transitive fanoutcone: y b c d e a b c d e a b c d e 4 3 1 1 3 1 1 4 6 6 4 3 1 1 a n1 1 1 2 4

  6. Previous work(2/2) 2 2 2 2 1 1 1 1 4 3 1 1 1 1 1 1 4 f f 10 • Case 3: The target wire is critical, and we rectify it at the transitive fanincone: a b c d e 6 4 3 1 1 a b a c a d a e

  7. Minimize the implementation cost • The sum of all weights and threshold values has been advocated by • VLSI Complexity of Discrete Neural Networks • Area-time performances of some neuralcomputations • Neural Information Processing Systems • Cost function

  8. input1 Choose the target wire(1/3) f T- w1 T • The target wire is critical input2 input3 inputn input1 input2 input3 inputn w1 w2 W3 wn w2 W3 wn f 1 1 2 n1 … … …

  9. Choose the target wire(2/3) f 3 3 • The target wire is not critical • The input whose weight is equal to the threshold value a b c d e b c d e 3 2 1 1 1 2 1 1 1 a f 3 3 1 1 3 1 n1

  10. Choose the target wire(3/3) f T T • The target wire is not critical • The input whose weight is equal to the threshold value Inputn+1 inputk input1 inputn Inputn+1 inputk w1 Wn Wn+1 wk Wn+1 wk 1 … … input1 inputn … … … 1 1 1 … f … n1 …

  11. START Proposed flow Input: A threshold network weight = threshold value? Yes No The target wire is critical? Yes • Wire removal& Rectification • Remove the target wire • Rectification network construction • OR connection • Wire removal & Rectification • Remove the target wire • Threshold value change • Rectification network construction • AND connection No Simplification Output: The threshold network with minimal implementation cost END

  12. Experimental results(1/2)

  13. Experimental results(2/2) T481 apex6 s13207 s9234

More Related