Monday, 22. May 2017


KTH Research Project Database

KTH Research Projec...
PROJECT




Architectural and Civil Engineering


Chemistry and biotechnology


Computer Science & Information Technology


Electrical Engineering


Mathematics


Physics and Energy


Production and Mechanical Engineering


Social Sciences and Humanities


Environment and sustainable development


OVERVIEWS

»Research Fields

»Organisational Structure

»Persons



Brain-like algorithms for temporal sequence processing - recognition, generation, and learning (»Add to Infobox)

Research Leader: Professor Anders Lansner


Computational Biology


Show picture in detail
The proposed project aims to extend an existing systems level multi-stage model of the mammalian neocortex with capabilities for spatio-temporal pattern learning, recognition, and associative memor...
The proposed project aims to extend an existing systems level multi-stage model of the mammalian neocortex with capabilities for spatio-temporal pattern learning, recognition, and associative memory. The existing model is an algorithmic realization of Hebb´s theory of cell assemblies and mathematical formalizations thereof in the form of sparsely connected and activated attractor associative memory. It represents cortical holistic perceptual processing based on the lateral long-range connectivity in neorcortical layers 2/3, together with mechanisms to generate a multi-stage cortical architecture based on self-organization and competitive learning, representing the local feature extraction capabilities mainly in neocortical layer 4. This allows building by activity dependent mechanisms a multi-stage cortical processing pipe-line with feed-forward, lateral, and feedback processing streams. This model will be extended with a broad range of conduction delays, thus providing it with the ability to capture temporal dependencies in the data, to build spatio-temporal feature detectors and to bind these together via delayed long-range connectivity. The extended model will be parallelized in a scalable fashion and benchmarked against experimental data on human sequence learning as well as by applying it for analysis of a large brain imaging (electroencephalography (EEG) and magnetoencephalography (MEG)) datasets, challenging the models spatio-temporal data analysis capabilities.

Period: 2013-01-01 - 2016-12-31

Keywords:
computational neuroscience, cortical associative memory, spiking neural network, Brain-IT


SEARCH



»Advanced Search



INFOBOX (0/0) show all
Your infobox is empty




Powered by AVERISinfo@avedas.com