Lectures notes.
Wasserman, "Advanced methods in neural computing", Reinhold New York 1993.
Learning Objectives
Educational Objectives:
The course aims to provide a deep knowledge of advanced computing techniques, used for data analysis of complex systems (es High Energy Physics). These models are inspired to “neural network” and try to reproduce some aspects of animal brain ability and its extraordinary efficiency of recognizing aggregated signals.
Acquired Competence:
Detailed knowledge of different models of artificial neural networks and their applications.
Acquired Skills (at the end of the course):
Construction of neural networks useful for data analysis using MATLAB or MATHEMATICA.
Type of Assessment
Oral exam
Course program
Neural dynamics of animal brain. Perceptron. Optimum Bayes classifier. Multilayer Perceptron. Learning algorithms. Radial Basis Functions. Clustering and multi-scales radial basis functions. Analysis of patterns in time. Widrow-Hoff’s filters. Non-linear control problems. Kohonen network. Principal Component Analysis. Statistical mechanics method: Hopefield network and recursive neural networks. Basic concepts of Statistical Mechanics. Spin glasses and Hopfield model. Replica simmetry. Parisi solution (Replica Symmetry Breaking). Ultrametricity and neural network; ierarchical organization in a disordered system. Optimization: Simulated Annealing and Monte Carlo methods. Genetic algorithms.
Toolbox of MATLAB.