neural network


Published on

abt neural network & it's application for seminar

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

neural network

  1. 1. Neural Network<br />Guided by<br />Chapala Maharana<br />Submitted By-<br />Nilmani Singh<br />Branch-CSE<br />Regd. No.-0601289251<br />
  2. 2. Contents<br />.Introduction to Neural Network<br />.Neurons<br />.Activation Function<br />.Types of Neural Network<br />.Learning In Neural Networks<br />.Application of Neural Network<br />.Advantages of Neural Network<br />.Disadvantages Of Neural Network<br />
  3. 3. Neural Networks<br /><ul><li>A method of computing, based on the interaction of multiple connected processing elements.
  4. 4. A powerful technique to solve many real world problems.
  5. 5. The ability to learn from experience in order to improve their performance.
  6. 6. Ability to deal with incomplete information</li></li></ul><li>Basics Of Neural Network<br /><ul><li>Biological approach to AI
  7. 7. Developed in 1943
  8. 8. Comprised of one or more layers of neurons
  9. 9. Several types, we’ll focus on feed-forward and feedback networks</li></li></ul><li>Neurons<br />Biological<br />Artificial<br />
  10. 10. Neural Network Neurons<br />Receives n-inputs<br />Multiplies each input by its weight<br />Applies activation function to the sum of results<br />Outputs result<br />
  11. 11. Activation Functions<br />Controls when unit is “active” or “inactive”<br />Threshold function outputs 1 when input is positive and 0 otherwise<br />Sigmoid function = 1 / (1 + e-x)<br />
  12. 12. Types of Neural Networks<br />Neural Network types can be classified based on following attributes:<br /><ul><li>Connection Type</li></ul> - Static (feedforward) - Dynamic (feedback)<br /><ul><li> Topology </li></ul> - Single layer - Multilayer - Recurrent<br /><ul><li> Learning Methods</li></ul> - Supervised - Unsupervised <br />- Reinforcement<br />
  13. 13. Classification Based On Connection Types<br /><ul><li>Static(Feedforward)
  14. 14. Dynamic(Feedback)</li></ul>Classification Based On Topology <br /><ul><li>Recurrent
  15. 15. Single layer
  16. 16. Multilayer</li></ul>(unit delay operator z-1implies dynamic system)<br />
  17. 17. Classification Based On Learning Method<br />Supervised<br />Unsupervised<br />Reinforcement<br />Supervised learning<br /><ul><li> Each training pattern: input + desired output
  18. 18. At each presentation: adapt weights
  19. 19. After many epochs convergence to a local minimum</li></li></ul><li>Unsupervised Learning<br />No help from the outside<br />No training data, no information available on the desired output<br />Learning by doing<br />Used to pick out structure in the input:<br />Clustering<br />Reduction of dimensionality  compression<br />Example: Kohonen’s Learning Law<br />
  20. 20. Reinforcement learning<br /><ul><li>Teacher: training data
  21. 21. The teacher scores the performance of the training examples
  22. 22. Use performance score to shuffle weights ‘randomly’
  23. 23. Relatively slow learning due to ‘randomness’</li></li></ul><li>Neural Network Applications<br /><ul><li>Pattern recognition
  24. 24. Investment analysis
  25. 25. Control systems & monitoring
  26. 26. Mobile computing
  27. 27. Marketing and financial applications
  28. 28. Forecasting – sales, market research, meteorology</li></li></ul><li>Advantages:<br /><ul><li>A neural network can perform tasks that a linear program can not.
  29. 29. When an element of the neural network fails, it can continue without any problem by their parallel nature.
  30. 30. A neural network learns and does not need to be reprogrammed.
  31. 31. It can be implemented in any application.
  32. 32. It can be implemented without any problem</li></li></ul><li>Disadvantages:<br /><ul><li>The neural network needs training to operate.
  33. 33. The architecture of a neural network is different from the architecture of microprocessors therefore needs to be emulated.
  34. 34. Requires high processing time for large neural networks.</li></li></ul><li>Conclusions<br />Neural networks provide ability to provide more human-like AI<br />Takes rough approximation and hard-coded reactions out of AI design (i.e. Rules and FSMs)<br />Still require a lot of fine-tuning during development<br />
  35. 35. References<br />Neural Networks, Fuzzy Logic, and Genetic Algorithm ( synthesis and Application) S.Rajasekaran, G.A. VijayalakshmiPai, PHI <br />Neuro Fuzzy and Soft Computing, J. S. R. JANG,C.T. Sun, E. Mitzutani, PHI<br />Neural Netware, a tutorial on neural networks<br />Sweetser, Penny. “Strategic Decision-Making with Neural Networks and Influence Maps”, AI Game Programming Wisdom 2, Section 7.7 (439 – 46)<br />Russell, Stuart and Norvig, Peter. Artificial Intelligence: A Modern Approach, Section 20.5 (736 – 48)<br />
  36. 36. THANK YOU<br />
  37. 37. QUESTIONS ?<br />