uu.seUppsala University Publications
Change search
ReferencesLink to record
Permanent link

Direct link
An Adaptive Sequential Monte Carlo Approach to Neural Network Training
Uppsala University, Disciplinary Domain of Humanities and Social Sciences, Faculty of Social Sciences, Department of Statistics.
2014 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

Sequential Monte Carlo methods (particle filter) have been successfully applied to the online training of neural networks. However the generic particle filter requires that model noise to be known prior to training. Furthermore, the random walk assumption with which the network weights are modeled by may be problematic as a result of the insufficient knowledge of the model noise. In this thesis, the evolution of the network weights are modeled using the Polynomial Prediction Model (PPM) which has been shown to have more predictive power than the random walk. The PPM can generate a whole class of models which can then be used in a modified multi-model version of the particle filter based on the Interacting Multiple Model (IMM) to train the neural network. The resulting algorithm generates an estimate of the noise terms which is closer to the true noise in the form of a weighted linear combination of the model noise given in the different models. This means that the algorithm can adapt to unknown model noise. Experiments show that the proposed algorithm offers better performance in training neural networks in the context where we are unable to determine the error terms.

Place, publisher, year, edition, pages
Keyword [en]
Neural Network, Particle Filter, Interacting Multiple Model (IMM), Polynomial Prediction Model (PPM)
National Category
Probability Theory and Statistics
URN: urn:nbn:se:uu:diva-225945OAI: oai:DiVA.org:uu-225945DiVA: diva2:722950
Subject / course
Educational program
Master Programme in Statistics
2014-06-03, F332, Ekonomikum, KyrkogÄrdsg 10, Uppsala, 11:10
Available from: 2014-06-24 Created: 2014-06-09 Last updated: 2014-06-24Bibliographically approved

Open Access in DiVA

No full text

By organisation
Department of Statistics
Probability Theory and Statistics

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: 306 hits
ReferencesLink to record
Permanent link

Direct link