Novytskyy D. Algebraic Methods for Theory of Neural Associative Memory

3Object of research: Artificial neural networks and their mathematical models The purpose of research: improvement of associative memory models and development of novel, more efficient learning algorithms for associative neural networks. Method of research: algebra, Riemannian geometry, mathematical statistics, optimization, functional analysis, simulation modeling, programming. Theoretical and practical results and novelty: associative memory models were improved using methods of algebra and Riemannian geometry. Some experimental facts were explained theoretically. Novel, more efficient associative-memory algorithms were developed and used for applied tasks. The degree of an application: the methods are implemented in the software neural computer Neuroland and applied to chemical image recognition in the system of "electronic nose" Object of application: information technologies, systems of data mining and artificial intelligence.