A Bayesian data augmentation approach for learning deep models

Simple item page
dc.contributor.authorTran, T.
dc.contributor.authorPham, T.
dc.contributor.authorCarneiro, G.
dc.contributor.authorPalmer, L.
dc.contributor.authorReid, I.
dc.contributor.conferenceNIPS Foundation Inc (4 Dec 2017 - 9 Dec 2017 : Long Beach, CA)
dc.contributor.editorGuyon, I.
dc.contributor.editorLuxburg, U.V.
dc.contributor.editorBengio, S.
dc.contributor.editorWallach, H.
dc.contributor.editorFergus, R.
dc.contributor.editorVishwanathan, S.
dc.contributor.editorGarnett, R.
dc.date.issued2018
dc.description.abstractData augmentation is an essential part of the training process applied to deep learning models. The motivation is that a robust training process for deep learning models depends on large annotated datasets, which are expensive to be acquired, stored and processed. Therefore a reasonable alternative is to be able to automatically generate new annotated training samples using a process known as data augmentation. The dominant data augmentation approach in the field assumes that new training samples can be obtained via random geometric or appearance transformations applied to annotated training samples, but this is a strong assumption because it is unclear if this is a reliable generative model for producing new training samples. In this paper, we provide a novel Bayesian formulation to data augmentation, where new annotated training points are treated as missing variables and generated based on the distribution learned from the training set. For learning, we introduce a theoretically sound algorithm --- generalised Monte Carlo expectation maximisation, and demonstrate one possible implementation via an extension of the Generative Adversarial Network (GAN). Classification results on MNIST, CIFAR-10 and CIFAR-100 show the better performance of our proposed method compared to the current dominant data augmentation approach mentioned above --- the results also show that our approach produces better classification results than similar GAN models.
dc.description.statementofresponsibilityToan Tran, Trung Pham, Gustavo Carneiro, Lyle Palmer and Ian Reid
dc.identifier.citationAdvances in neural information processing systems, 2018 / Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (ed./s), vol.2017-December, pp.1-10
dc.identifier.issn1049-5258
dc.identifier.orcidTran, T. [0000-0001-7182-7548]
dc.identifier.orcidCarneiro, G. [0000-0002-5571-6220]
dc.identifier.orcidPalmer, L. [0000-0002-1628-3055]
dc.identifier.orcidReid, I. [0000-0001-7790-6423]
dc.identifier.urihttp://hdl.handle.net/2440/117310
dc.language.isoen
dc.publisherNeural Information Processing Systems Foundation
dc.relation.granthttp://purl.org/au-research/grants/arc/CE140100016
dc.relation.granthttp://purl.org/au-research/grants/arc/FL130100102
dc.relation.ispartofseriesAdvances in Neural Information Processing Systems
dc.rightsCopyright status unknown
dc.source.urihttps://papers.nips.cc/paper/6872-a-bayesian-data-augmentation-approach-for-learning-deep-models
dc.titleA Bayesian data augmentation approach for learning deep models
dc.typeConference paper
pubs.publication-statusPublished

Files

Collections

Computer Science publications
Australian Institute for Machine Learning publications