Laboratory of Mathematical Modeling

Application of developments in the creation of intelligent heterogeneous data processing systems for solving various applied problems in agriculture and medicine based on neural network analysis of signals and images.

Development of multimodal neural network architectures using heterogeneous data processing algorithms to improve the intelligent analysis of signals and images. Improvement of existing methods and algorithms of gradient-free optimization and gradient optimization based on fractional rational derivatives to accelerate and improve the training of deep neural networks.

Modification of the architectures of computing elements used in modern high-speed signal and image processing devices. Adaptation of the developed methods and architectures for high-speed parallel computing over several channels in a system of residual classes. Improving approaches to one-dimensional and multidimensional convolution by developing the Grape methodology for implementing matrix calculations with reduced computational complexity in signal and image processing.