|
 |
Module I
Introduction to Artificial neural Networks - elementary neurophysiology - Mc culloch and pitts model - Rosenblatts perceptron - perceptron convergence theorem (without proof) - LMS algorithms - linear separability - multilayer perceptrons - Derivation of backpropagation algorithm - practical considerations - approximation of function using MLPs - universal approximation theorem.
Module II
Hopefield networks - energy function - spurious states - error performance - storage capacity, simulated annealing, Boltzmann machine - Boltzmann learning rule, meanfield theory machine - MFT learning algorithms, self organisation - SOFM algorithms - Hierarchical vector quantisation, Adaptive resonance theory (ART) - ARTI clustering algorithm - ARTI network
Module III
Introduction to RBF networks - covers theorem on separability of pattern - RBF learning strategies - k means and LMS algorithms - comparison of RBF networks and MLPs.
Applications of neural networks - image compression using MLPs- character retrieval using discrete hopefield networks- Solution of travelling sales man problem using hopefield networks.
Text Book: -
Ref (1) 1
References: -
1. Simon Haykins - Neural networks, Macmillan College of publishing company 1994
2. Bose & Liang - neural network fundamentals - Mc Graw Hill - 1996
3. Hertz, Krogh and palmer - introduction to the theory of Neural computation, addison Wesley,1991
4. Zurada - introduction to Artificial Neural Networks, jaico publishing 1992
5. R P Liiman - An introduction to computing with neural networks, IEEE ASSP magazine, pp 4-22, April 1987
6. Hush & Horne - Progress in supervision neural networks, IEEE signal processing magazine, pp8-39, January 1993.
Question paper: -
The question paper will consist of two parts. Part 1 is to cover entire syllabus, and compulsory for 40 marks. This may contain 10 questions of 4marks each. Part II is to cover 3 modules. There can be 3 questions from each module ( 10 marks each) out of which 2 are to be answered.
|
|
|
