Multi-task and multi-kernel gaussian process dynamical systems

File Description SizeFormat 
Manuscript_stamped.pdfFile embargoed until 18 December 20173.53 MBAdobe PDF    Request a copy
Title: Multi-task and multi-kernel gaussian process dynamical systems
Author(s): Korkinof, D
Demiris, Y
Item Type: Journal Article
Abstract: In this work, we propose a novel method for rectifying damaged motion sequences in an unsupervised manner. In order to achieve maximal accuracy, the proposed model takes advantage of three key properties of the data: their sequential nature, the redundancy that manifests itself among repetitions of the same task, and the potential of knowledge transfer across different tasks. In order to do so, we formulate a factor model consisting of Gaussian Process Dynamical Systems (GPDS), where each factor corresponds to a single basic pattern in time and is able to represent their sequential nature. Factors collectively form a dictionary of fundamental trajectories shared among all sequences, thus able to capture recurrent patterns within the same or across different tasks. We employ variational inference to learn directly from incomplete sequences and perform maximum a-posteriori (MAP) estimates of the missing values. We have evaluated our model with a number of motion datasets, including robotic and human motion capture data. We have compared our approach to well-established methods in the literature in terms of their reconstruction error and our results indicate significant accuracy improvement across different datasets and missing data ratios. Concluding, we investigate the performance benefits of the multi-task learning scenario and how this improvement relates to the extent of component sharing that takes place.
Publication Date: 18-Dec-2016
Date of Acceptance: 14-Dec-2016
URI: http://hdl.handle.net/10044/1/43499
DOI: https://dx.doi.org/10.1016/j.patcog.2016.12.014
ISSN: 1873-5142
Publisher: Elsevier
Start Page: 190
End Page: 201
Journal / Book Title: Pattern Recognition
Volume: 66
Copyright Statement: © 2016, Elsevier Ltd. All rights reserved. This manuscript is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International http://creativecommons.org/licenses/by-nc-nd/4.0/
Sponsor/Funder: Commission of the European Communities
Funder's Grant Number: 612139
Keywords: Science & Technology
Technology
Computer Science, Artificial Intelligence
Engineering, Electrical & Electronic
Computer Science
Engineering
Gaussian processes
Variational Bayes
Matrix decomposition
Factor models
Data completion
Human motion
Gaussian process latent variable models
Multi-task learning
Unsupervised learning
MATRIX COMPLETION
FACTORIZATION
Artificial Intelligence & Image Processing
0899 Other Information And Computing Sciences
0906 Electrical And Electronic Engineering
0801 Artificial Intelligence And Image Processing
Publication Status: Published
Embargo Date: 2017-12-18
Appears in Collections:Faculty of Engineering
Electrical and Electronic Engineering



Items in Spiral are protected by copyright, with all rights reserved, unless otherwise indicated.

Creative Commons