2026-09-19T19:54:10Zhttps://dspace.mit.edu/server/oai/request

oai:dspace.mit.edu:1721.1/1299052026-06-06T00:55:33Zcom_1721.1_7582com_1721.1_7581col_1721.1_131023

Tomaso Poggio. Henry, Timothy G. Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science. Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science 2021-02-19T20:49:40Z 2021-02-19T20:49:40Z 2020 2020 https://hdl.handle.net/1721.1/129905 1237411492 Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, February, 2020 Cataloged from student-submitted PDF of thesis. Includes bibliographical references (pages 71-73). Visual understanding results from a combined understanding of primitive visual attributes such as color, texture, and shape. This allows humans and other primates to generalize their understanding of objects to new combinations of attributes. For instance, one can understand that a pink elephant is an elephant even if they have never seen this particular combination of color and shape before. However, is it the case that deep neural networks (DNNs) are able to generalize to such novel combinations in object recognition or other related vision tasks? This thesis demonstrates that (1) the ability of DNNs to generalize to unseen attribute combinations increases with the increased diversity of combinations seen in training as a percentage of the total combination space, (2) this effect is largely independent of the specifics of the DNN architecture used, (3) while single-task and multi-task formulations of supervised attribute classification problems may lead to similar performance on seen combinations, single-task formulations have a superior ability to generalize to unseen combinations, and (4) DNNs demonstrating the ability to generalize well in this setting learn to do so by leveraging emergent hidden units that exhibit properties of attribute selectivity and invariance. by Timothy G. Henry. M.Eng. M.Eng. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science 73 pages eng Massachusetts Institute of Technology MIT theses may be protected by copyright. Please reuse MIT thesis content according to the MIT Libraries Permissions Policy, which is available through the URL provided. http://dspace.mit.edu/handle/1721.1/7582 Electrical Engineering and Computer Science. Generalization of deep neural networks to unseen attribute combinations Thesis application/pdf 2021-02-19T20:49:10Z Publication Master EECS unknown unknown <Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="b3771ebf-78f9-4ae2-935f-511afe7ea502"> <Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843</Type> <Language>eng</Language> <Title>Generalization of deep neural networks to unseen attribute combinations</Title> <PublishedIn> <Publication> </Publication> </PublishedIn> <PublicationDate>2020</PublicationDate> <Authors> <Author> <DisplayName>Henry, Timothy G.</DisplayName> <Affiliation> <OrgUnit> </OrgUnit> </Affiliation> </Author> </Authors> <Editors> </Editors> <Publishers> <Publisher> <DisplayName>Massachusetts Institute of Technology</DisplayName> <OrgUnit /> </Publisher> </Publishers> <License>http://dspace.mit.edu/handle/1721.1/7582</License> <Keyword>Electrical Engineering and Computer Science.</Keyword> <Abstract>Visual understanding results from a combined understanding of primitive visual attributes such as color, texture, and shape. This allows humans and other primates to generalize their understanding of objects to new combinations of attributes. For instance, one can understand that a pink elephant is an elephant even if they have never seen this particular combination of color and shape before. However, is it the case that deep neural networks (DNNs) are able to generalize to such novel combinations in object recognition or other related vision tasks? This thesis demonstrates that (1) the ability of DNNs to generalize to unseen attribute combinations increases with the increased diversity of combinations seen in training as a percentage of the total combination space, (2) this effect is largely independent of the specifics of the DNN architecture used, (3) while single-task and multi-task formulations of supervised attribute classification problems may lead to similar performance on seen combinations, single-task formulations have a superior ability to generalize to unseen combinations, and (4) DNNs demonstrating the ability to generalize well in this setting learn to do so by leveraging emergent hidden units that exhibit properties of attribute selectivity and invariance.</Abstract> <Access xmlns="http://purl.org/coar/access_right" > </Access> </Publication>