2026-09-20T04:26:16Zhttps://dspace.mit.edu/server/oai/request

oai:dspace.mit.edu:1721.1/1216252026-06-06T00:54:47Zcom_1721.1_7582com_1721.1_7581col_1721.1_131023

Daniel J. Weitzner. Dethy, Elizabeth(Elizabeth A.) Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science. Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science 2019-07-15T20:29:01Z 2019-07-15T20:29:01Z 2018 2018 https://hdl.handle.net/1721.1/121625 1098171971 This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2018 Cataloged from student-submitted PDF version of thesis. Includes bibliographical references (pages 55-56). In this thesis I test for the emergence of policy relevant features as disentangled, human interpretable representations in facial characterization networks. I probe an age and gender classifier using the NetworkDissection method with a hand labelled dataset of facial features, skin tones, and textures. Facial features and skin tones emerge as disentangled concepts in each of the networks probed. The emergence of these features in a smaller image classification network indicates the effectiveness of the NetworkDissection method in contexts other than ones studied in the original paper. Moreover, the emergence of policy relevant features, skin tones, indicates the method may be effective in identifying policy sensitive attributes. I also analyze the robustness of the NetworkDissection technique itself to changes in a key component of the experimental setup: the source of ground truth human understandable concepts. The results demonstrate the technique reliably applies labels when new concepts and samples are added to the set of ground truth labels. by Elizabeth Dethy. M.Eng. M.Eng. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science 56 pages eng Massachusetts Institute of Technology MIT theses are protected by copyright. They may be viewed, downloaded, or printed from this source but further reproduction or distribution in any format is prohibited without written permission. http://dspace.mit.edu/handle/1721.1/7582 Electrical Engineering and Computer Science. Assessing the usefulness of NetworkDissection in identifying the interpretability of facial characterization networks Thesis application/pdf 2019-07-15T20:28:58Z Publication Master EECS unknown unknown <Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="77d2ef91-8361-463a-a633-d717df15345f"> <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>Assessing the usefulness of NetworkDissection in identifying the interpretability of facial characterization networks</Title> <PublishedIn> <Publication> </Publication> </PublishedIn> <PublicationDate>2018</PublicationDate> <Authors> <Author> <DisplayName>Dethy, Elizabeth(Elizabeth A.)</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>In this thesis I test for the emergence of policy relevant features as disentangled, human interpretable representations in facial characterization networks. I probe an age and gender classifier using the NetworkDissection method with a hand labelled dataset of facial features, skin tones, and textures. Facial features and skin tones emerge as disentangled concepts in each of the networks probed. The emergence of these features in a smaller image classification network indicates the effectiveness of the NetworkDissection method in contexts other than ones studied in the original paper. Moreover, the emergence of policy relevant features, skin tones, indicates the method may be effective in identifying policy sensitive attributes. I also analyze the robustness of the NetworkDissection technique itself to changes in a key component of the experimental setup: the source of ground truth human understandable concepts. The results demonstrate the technique reliably applies labels when new concepts and samples are added to the set of ground truth labels.</Abstract> <Access xmlns="http://purl.org/coar/access_right" > </Access> </Publication>