{"id":{"repo_id":"mit","oai_identifier":"oai:dspace.mit.edu:1721.1/139151"},"canonical_url":"https://search.dev.ndltd.org/etd/mit/oai:dspace.mit.edu:1721.1/139151","repository":{"repo_id":"mit","name":"MIT","base_url":"https://dspace.mit.edu/oai/request"},"display":{"title":"Rewriting the Rules of a Classifier","abstract":"Observations of various deep neural network architectures indicate that deep networks may be spontaneously learning representations of concepts with semantic meaning, and encoding a relational structure or rule between these concepts. We refer to these encoded relationships between concepts in the network as rules. In classifiers, we rewrite an existing rule in the network as desired, referred to as the rewriting technique. We demonstrate that using our rewriting technique and simple human knowledge about how to classify the world around us, we can generalize existing classes to unseen variants, identify spurious correlations present in the dataset, mitigate the effects of spurious correlations, and introduce new classes. We find that our technique reduces the need for: computing resources, because we only re-train a single layer’s weights; new training images, because our rewriting technique can rewrite using concepts already encoded in the network; and domain knowledge, because what we choose to edit to improve classification is derived from logical rules a human would construct to classify images.","abstract_html":"Observations of various deep neural network architectures indicate that deep networks may be spontaneously learning representations of concepts with semantic meaning, and encoding a relational structure or rule between these concepts. We refer to these encoded relationships between concepts in the network as rules. In classifiers, we rewrite an existing rule in the network as desired, referred to as the rewriting technique. We demonstrate that using our rewriting technique and simple human knowledge about how to classify the world around us, we can generalize existing classes to unseen variants, identify spurious correlations present in the dataset, mitigate the effects of spurious correlations, and introduce new classes. We find that our technique reduces the need for: computing resources, because we only re-train a single layer’s weights; new training images, because our rewriting technique can rewrite using concepts already encoded in the network; and domain knowledge, because what we choose to edit to improve classification is derived from logical rules a human would construct to classify images.","abstract_has_math":false,"creators":["Elango, Mahalaxmi"],"institution":"Massachusetts Institute of Technology","degree_name":"Master","degree_level":null,"degree_discipline":null,"degree_department":"Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science","school":null,"contributors":[],"advisors":["Torralba, Antonio"],"committee_chairs":[],"committee_members":[],"year":2021,"date_issued":"2021-06","date_published":"2021-06","updated_at":"2026-07-22T22:22:26Z","subjects":[],"languages":[],"rights":["In Copyright - Educational Use Permitted","Copyright MIT"],"rights_urls":["http://rightsstatements.org/page/InC-EDU/1.0/"],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1721.1/139151","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Torralba, Antonio"]},{"key":"dc:contributor.department","label":"Department","values":["Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science"]},{"key":"dc:creator","label":"Author","values":["Elango, Mahalaxmi"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2022-01-14T14:52:58Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2022-01-14T14:52:58Z"]},{"key":"dc:date.issued","label":"Date","values":["2021-06"]},{"key":"dc:publisher","label":"Institution","values":["Massachusetts Institute of Technology"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master","Master of Engineering in Electrical Engineering and Computer Science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["In Copyright - Educational Use Permitted","Copyright MIT"]},{"key":"dc:rights.uri","label":"Rights URI","values":["http://rightsstatements.org/page/InC-EDU/1.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1721.1/139151"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Observations of various deep neural network architectures indicate that deep networks may be spontaneously learning representations of concepts with semantic meaning, and encoding a relational structure or rule between these concepts. We refer to these encoded relationships between concepts in the network as rules. In classifiers, we rewrite an existing rule in the network as desired, referred to as the rewriting technique. We demonstrate that using our rewriting technique and simple human knowledge about how to classify the world around us, we can generalize existing classes to unseen variants, identify spurious correlations present in the dataset, mitigate the effects of spurious correlations, and introduce new classes. We find that our technique reduces the need for: computing resources, because we only re-train a single layer’s weights; new training images, because our rewriting technique can rewrite using concepts already encoded in the network; and domain knowledge, because what we choose to edit to improve classification is derived from logical rules a human would construct to classify images."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["M.Eng."]},{"key":"dc:title","label":"Title","values":["Rewriting the Rules of a Classifier"]}]}],"canonical_facts":{"dc:contributor.advisor":["Torralba, Antonio"],"dc:contributor.department":["Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science"],"dc:creator":["Elango, Mahalaxmi"],"dc:date.accessioned":["2022-01-14T14:52:58Z"],"dc:date.available":["2022-01-14T14:52:58Z"],"dc:date.issued":["2021-06"],"dc:description.abstract":["Observations of various deep neural network architectures indicate that deep networks may be spontaneously learning representations of concepts with semantic meaning, and encoding a relational structure or rule between these concepts. We refer to these encoded relationships between concepts in the network as rules. In classifiers, we rewrite an existing rule in the network as desired, referred to as the rewriting technique. We demonstrate that using our rewriting technique and simple human knowledge about how to classify the world around us, we can generalize existing classes to unseen variants, identify spurious correlations present in the dataset, mitigate the effects of spurious correlations, and introduce new classes. We find that our technique reduces the need for: computing resources, because we only re-train a single layer’s weights; new training images, because our rewriting technique can rewrite using concepts already encoded in the network; and domain knowledge, because what we choose to edit to improve classification is derived from logical rules a human would construct to classify images."],"dc:description.degree":["M.Eng."],"dc:identifier.uri":["https://hdl.handle.net/1721.1/139151"],"dc:publisher":["Massachusetts Institute of Technology"],"dc:rights":["In Copyright - Educational Use Permitted","Copyright MIT"],"dc:rights.uri":["http://rightsstatements.org/page/InC-EDU/1.0/"],"dc:title":["Rewriting the Rules of a Classifier"],"dc:type":["Thesis"],"thesis:degree_name":["Master","Master of Engineering in Electrical Engineering and Computer Science"]},"updated_at":"2026-07-22T22:22:26Z"}