Nis1

820 views
687 views

Published on

Published in: Technology, Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
820
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
26
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Nis1

  1. 1. Soft computing (SC) <ul><li>Objective: </li></ul><ul><li>Mimic human (linguistic) reasoning </li></ul><ul><li>Main constituents: </li></ul><ul><li>- Fuzzy systems </li></ul><ul><li>- Neural networks </li></ul><ul><li>- Evolutionary computing </li></ul><ul><li>- Probabilistic reasoning </li></ul>VAN-00
  2. 2. Constituents of SC <ul><li>Fuzzy systems => imprecision </li></ul><ul><li>Neural networks => learning </li></ul><ul><li>Probabilistic reasoning => uncertainty </li></ul><ul><li>Evolutionary computing => optimization </li></ul>VAN-00 Over 24 000 publications today
  3. 3. SC: a user-friendly interface VAN-00
  4. 4. Advantages of SC <ul><li>Models base on human reasoning. </li></ul><ul><li>Models can be - linguistic - simple (no number crunching), - comprehensible (no black boxes), - fast when computing, - good in practice. </li></ul>VAN-00
  5. 5. SC today (Zadeh) <ul><li>Computing with words (CW) </li></ul><ul><li>Theory of information granulation (TFIG) </li></ul><ul><li>Computational theory of perceptions (CTP) </li></ul>VAN-00
  6. 6. Possible SC data & operations <ul><li>Numeric data: 5, about 5, 5 to 6, about 5 to 6 </li></ul><ul><li>Linguistic data: cheap, very big, not high, medium or bad </li></ul><ul><li>Functions & relations: f(x), about f(x), fairly similar, much greater </li></ul>VAN-00
  7. 7. Neural networks (NN, 1940's) <ul><li>Neural networks offer a powerful method to explore, classify, and identify patterns in data. </li></ul><ul><li>Website of Matlab </li></ul><ul><li>Neuron: y=  w i x i </li></ul>VAN-00
  8. 8. Machine learning (supervised) <ul><li>Pattern recognition based on training data. </li></ul><ul><li>Classification supervised by instructor. </li></ul><ul><li>Neural (crisp or fuzzy), neuro-fuzzy and fuzzy models. </li></ul>VAN-00 Peach Plum ? Instructor
  9. 9. Machine learning (unsupervised) <ul><li>Pattern recognition based on training data. </li></ul><ul><li>Classification based on structure of data (clustering). </li></ul><ul><li>Neural (crisp or fuzzy), neuro-fuzzy and fuzzy models. </li></ul>VAN-00 Peach Plum Nectarine Labeling
  10. 10. Machine learning (unsupervised) <ul><li>Self-organized maps (Kohonen). </li></ul><ul><li>Fuzzy c-means (Bezdek). </li></ul><ul><li>Subclust (Yager, Chiu). </li></ul>VAN-00 Peach Plum Nectarine Labeling Websom Self-Organizing Maps for Internet Exploration
  11. 11. Fuzzy systems (Zadeh, 1960's) <ul><li>Deal with imprecise entities in automated environments (computer environments) </li></ul><ul><li>Base on fuzzy set theory and fuzzy logic. </li></ul><ul><li>Most applications in control and decision making </li></ul>VAN-00 Omron’s fuzzy processor Omron Electronics Matlab's Fuzzy Logic Toolbox
  12. 12. SC applications: control <ul><li>Heavy industry (Matsushita, Siemens, Stora-Enso) </li></ul><ul><li>Home appliances (Canon, Sony, Goldstar, Siemens) </li></ul><ul><li>Automobiles (Nissan, Mitsubishi, Daimler-Chrysler, BMW, Volkswagen) </li></ul><ul><li>Spacecrafts (NASA) </li></ul>VAN-00
  13. 13. SC applications: business VAN-00 <ul><li>hospital stay prediction, </li></ul><ul><li>TV commercial slot evaluation, </li></ul><ul><li>address matching, </li></ul><ul><li>fuzzy cluster analysis, </li></ul><ul><li>sales prognosis for mail order house, </li></ul><ul><li>multi-criteria optimization etc. </li></ul><ul><li>(source: FuzzyTech) </li></ul><ul><li>supplier evaluation for sample testing, </li></ul><ul><li>customer targeting, </li></ul><ul><li>sequencing, </li></ul><ul><li>scheduling, </li></ul><ul><li>optimizing R&D </li></ul><ul><li>projects, </li></ul><ul><li>knowledge-based prognosis, </li></ul><ul><li>fuzzy data analysis </li></ul>
  14. 14. SC applications: finance <ul><li>Fuzzy scoring for mortgage applicants, </li></ul><ul><li>creditworthiness assessment, </li></ul><ul><li>fuzzy-enhanced score card for lease risk assessment, </li></ul><ul><li>risk profile analysis, </li></ul><ul><li>insurance fraud detection, </li></ul><ul><li>cash supply optimization, </li></ul><ul><li>foreign exchange trading, </li></ul><ul><li>insider </li></ul><ul><li>trading surveillance, </li></ul><ul><li>investor classification etc. </li></ul><ul><li>Source: FuzzyTech </li></ul>VAN-00
  15. 15. SC applications: robotics VAN-00 Fukuda’s lab Joseph F. Engelberger We are proud to announce that the HelpMate Robotic Courier has been acquired by Pyxis Corporation . Entertainment robot AIBO
  16. 16. SC applications: others VAN-00 <ul><li>Statistics </li></ul><ul><li>Social sciences </li></ul><ul><li>Behavioural sciences </li></ul><ul><li>Biology </li></ul><ul><li>Medicine </li></ul>
  17. 17. (Neuro)-fuzzy system construction VAN-00 Training data Experts Fuzzy rules (SOM, c-means etc.) Control data System evaluation (errors) Tuning (NN) New system
  18. 18. Model construction (mathematical) <ul><li>Mathematical models are functions. Deep knowledge on mathematics. </li></ul><ul><li>If non-linear (eg. NN), laborious calculations and computing. </li></ul><ul><li>Linear models can be too simplified. </li></ul><ul><li>How can we find appropriate functions? </li></ul>VAN-00 Y=1-1./(1 + EXP(-2*(X-5)))
  19. 19. Model construction (trad. rules ) VAN-00 If 0 < x<1, then y=1 If 1 < x<2, then y=0.99 : If 8 < x < 10, then y=0 If 0 < x<1, then y=f(x) If 1 < x<2, then y=g(x) : If 8 < x < 10, then y=h(x) - Rule for each input. => Large rule bases. - Only one rule is fired for each input. - Coarse models.
  20. 20. Model construction (SC/fuzzy) VAN-00 If x  0, then y  1 If x  5, then y  0.5 If x  10, then y  0 - Approximate values - Rules only describe typical cases (no rule for each input). => Small rule bases. - A group of rules are partially fired simultaneously.
  21. 21. SC and future <ul><li>SC and conventional methods should be used in combination. </li></ul>VAN-00

×