Deep learning methods for likelihood-free inference :approximating the posterior distribution with convolutional neural networks

Mertens, Ulf Kai

[thumbnail of Dissertation_A_Ulf_Mertens.pdf]

PDF, English - main document
Download (4MB) | Terms of use

Citation of documents: Please do not cite the URL that is displayed in your browser location input, instead use the DOI, URN or the persistent URL below, as we can guarantee their long-time accessibility.

DOI: 10.11588/heidok.00026383
URN: urn:nbn:de:bsz:16-heidok-263839

Abstract

In this thesis, a fast and likelihood-free approach for parameter inference is introduced. The convolutional neural network, named DeepInference, learns to predict the posterior mean and variance of multi-dimensional posterior distributions from raw simulated data. It is shown how DeepInference can be applied to the drift diffusion model (DDM) and the Lévy flight model, a likelihood-free extension of the DDM. For both models, state-of-the-art results in terms of accuracy of parameter estimation are observed.

Document type:	Dissertation
Supervisor:	Voß, Prof. Dr. Andreas
Place of Publication:	Heidelberg, Deutschland
Date of thesis defense:	11 April 2019
Date Deposited:	10 May 2019 11:46
Date:	2019
Faculties / Institutes:	The Faculty of Behavioural and Cultural Studies > Institute of Psychology
DDC-classification:	004 Data processing Computer science 150 Psychology 310 General statistics
Controlled Keywords:	Likelihood, Maschinelles Lernen, Neuronales Netz
Uncontrolled Keywords:	Deep learning; convolutional neural network; Approximate Bayesian Computation