Change search
ReferencesLink to record
Permanent link

Direct link
Cortex Inspired Network Architectures For Spatio-Temporal Information Processing
KTH, School of Computer Science and Communication (CSC), Computational Biology, CB. (Lansner lab)
2013 (English)Independent thesis Advanced level (degree of Master (Two Years)), 80 credits / 120 HE creditsStudent thesis
Abstract [en]

The abundance of high-dimensional datasets providesscientists with a strong foundation in their research. With high-performancecomputing platforms becoming increasingly available and more powerful,large-scale data processing represents an important step toward modeling andunderstanding the underlying processes behind such data.

In this thesis, we propose a general cortex-inspired information processingnetwork architecture capable of capturing spatio-temporal correlations in dataand forming distributed representations as cortical activation patterns. Theproposed architecture has a modular and multi-layered organization which isefficiently parallelized to allow large-scale computations. The network allowsunsupervised processing of multivariate stochastic time series, irregardless ofthe data source, producing a sparse de-correlated representation of the inputfeatures expanded by time delays.

The features extracted by the architecture are then used for supervised learningwith Bayesian confidence propagation neural networks and evaluated on speechclassification and recognition tasks. Due to their rich temporal dynamics, weexploited auditory signals for speech recognition as an use case for performanceevaluation. In terms of classification performance, the proposed architectureoutperforms modern machine-learning methods such as support vector machines andobtains results comparable to other state-of-the-art speech recognition methods.The potential of the proposed scalable cortex-inspired approach to capturemeaningful multivariate temporal correlations and provide insight into themodel-free high- dimensional data decomposition basis is expected to be ofparticular use in the analysis of large brain signal datasets such as EEG orMEG.

Place, publisher, year, edition, pages
2013. , 101 p.
Keyword [en]
cortical models, large-scale data analysis, spatio-temporal recognition, speech recognition, brain signal processing
National Category
Computer Science
URN: urn:nbn:se:kth:diva-129758OAI: diva2:678507
Subject / course
Computer Science
Educational program
Master of Science - Systems Biology
2013-08-22, RB35, Roslagstullsbacken 35, Stockholm, 15:15 (English)
Available from: 2013-12-13 Created: 2013-10-04 Last updated: 2013-12-13Bibliographically approved

Open Access in DiVA

No full text

Search in DiVA

By author/editor
Franović, Tin
By organisation
Computational Biology, CB
Computer Science

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 268 hits
ReferencesLink to record
Permanent link

Direct link