ARTIFICIAL INTELLIGENCE AND KNOWLEDGE BASED SYSTEMS:
(12 hours) (KW)
[1-2] Historical Overview: Definition of Artificial Intelligence (AI). Application areas. General problem solving vs specific knowledge. Complexity.
[3-6] Heuristic Search: Uninformed vs informed search strategies. Formal properties A*. Minimax game search, alpha-beta pruning.
[7-10] Logic and Resolution: Knowledge Representation, Propositional and Predicate Calculus. Inference Rules. Clause Form. Resolution Strategies.
[11-12] Case studies: Applications in medicine.
PATTERN RECOGNITION: (12 hours) (JVK)
 Introduction. The pattern recognition problem. Model of pattern recognition system. Statistical and structural approach. Basic terminology.
[2-3] Decision Theory. Review of mathematical prerequisites and probability theory. Elements of statistical decision theory. Bayes decision rule. Parametric decision rule for normally distributed classes.
[4-5] Classifiers. Discriminant functions. Support Vector Machines. Nearest neighbour decision rules. Error estimation. Multiple classifiers.
[6-7] Non-supervised Learning. Cluster analysis. Concept of a cluster. Cluster models. Similarity measures. Clustering criteria and algorithms.
[8-9] Feature Selection and Extraction. Class separability measures. Feature set search algorithms (branch and bound, sub-optimal algorithms). Feature extraction. Karhunen-Loeve transform.
[10-12] Case Studies. One of the following: Face recognition. Detection of microcalcification in digital mammograms. Video segmentation. Speech recognition.
NEURAL NETWORKS: (6 hours) (TW)
 Artificial NN models: biological plausibility, feedback concepts, network architecture overview, discriminate functions, learning paradigms, linear associative memories.
[2-3] Single-Layer feed forward: discrete and continuous perceptrons, convergence theorem, comparison with Gaussian classifier, nonseparability, LMS algorithms, steepest descent.
[4-5] Multilayer Perceptrons: Learning in multilayer perceptrons, target function, objective functions, parameters of learning process, convergence, network growing and pruning.
 Case studies: Applications of neural network techniques.