May 21, 2019   8:35 p.m. Zina
Academic information system

Projects


Echo state neural networks

Supervisor: prof. RNDr. Jiří Pospíchal, DrSc.


Basic information   Workers      


This page shows details on the project. The primary projects are displayed together with a list of sub-projects.

Project description:The goal of the project is a study of a modern approach to recurrent neural networks, exceptionally suitable both for a time series prediction, as well as for modeling of cognitive processes in artificial neural systems. Neural network in this approach includes a block of neurons with a recurrent architecture, which is randomly generated and the weight coefficient of its connections are fixed during the learning stage of the network. The input activities incoming to the neural network will be mapped onto a rich dynamic structure of activities of hidden neurons, which are used as an input to the layer of output neurons. The proposed project will be focused on the study of generalizing properties of these networks and their application in modeling of cognitive processes with help of modular structures, while single modules are built from echo state neural networks.
Kind of project:VEGA ()
Department:Institute of Applied Informatics (FIIT)
Project identification:1/1047/04
Project status:Successfully completed
Project start date :01. 01. 2004
Project close date:31. 12. 2006
Number of workers in the project:1
Number of official workers in the project:0