Timezone: »
Distributed representations provide a vector space that captures meaningful relationships between data instances. The distributed nature of these representations, however, entangles together multiple attributes or concepts of data instances (e.g., the topic or sentiment of a text, characteristics of the author (age, gender, etc), etc). Recent work has proposed the task of concept erasure, in which rather than making a concept predictable, the goal is to remove an attribute from distributed representations while retaining other information from the original representation space as much as possible. In this paper, we propose a new distance metric learning-based objective, the Kernelized Rate-Distortion Maximizer (KRaM), for performing concept erasure. KRaM fits a transformation of representations to match a specified distance measure (defined by a labeled concept to erase) using a modified rate-distortion function. Specifically, KRaM's objective function aims to make instances with similar concept labels dissimilar in the learned representation space while retaining other information. We find that optimizing KRaM effectively erases various types of concepts—categorical, continuous, and vector-valued variables—from data representations across diverse domains. We also provide a theoretical analysis of several properties of KRaM's objective. To assess the quality of the learned representations, we propose an alignment score to evaluate their similarity with the original representation space. Additionally, we conduct experiments to showcase KRaM's efficacy in various settings, from erasing binary gender variables in word embeddings to vector-valued variables in GPT-3 representations.
Author Information
Somnath Basu Roy Chowdhury (Department of Computer Science, University of North Carolina, Chapel Hill)
Nicholas Monath (Google DeepMind)
Kumar Avinava Dubey (Google Research)
Amr Ahmed (Google Research)
Amr Ahmed is a Senior Staff Research Scientist at Google. He received his M.Sc and PhD degrees from the School of Computer Science, Carnegie Mellon University in 2009 and 2011, respectively. He received the best paper award at KDD 2014 , the best Paper Award at WSDM 2014, the 2012 ACM SIGKDD Doctoral Dissertation Award, and a best paper award (runner-up) at WSDM 2012. He co-chaired the WWW'18 track on Web Content Analysis and served as an Area Chair for IJCAI 2019, SIGIR 2019, SIGIR 2018, ICML 2018, ICML 2017, KDD 2016, WSDM 2015, ICML 2014, and ICDM 2014. His research interests include large-scale machine learning, data/web mining, user modeling, personalization, social networks and content analysis.
Snigdha Chaturvedi (Department of Computer Science, University of North Carolina, Chapel Hill)
More from the Same Authors
-
2023 Poster: Dense-Exponential Random Features: Sharp Positive Estimators of the Gaussian Kernel »
Valerii Likhosherstov · Krzysztof M Choromanski · Kumar Avinava Dubey · Frederick Liu · Tamas Sarlos · Adrian Weller -
2023 Poster: Mnemosyne: Learning to Train Transformers with Transformers »
Deepali Jain · Krzysztof M Choromanski · Kumar Avinava Dubey · Sumeet Singh · Vikas Sindhwani · Tingnan Zhang · Jie Tan -
2022 Poster: Chefs' Random Tables: Non-Trigonometric Random Features »
Valerii Likhosherstov · Krzysztof M Choromanski · Kumar Avinava Dubey · Frederick Liu · Tamas Sarlos · Adrian Weller -
2022 Poster: A Fourier Approach to Mixture Learning »
Mingda Qiao · Guru Guruganesh · Ankit Rawat · Kumar Avinava Dubey · Manzil Zaheer -
2020 Poster: Latent Bandits Revisited »
Joey Hong · Branislav Kveton · Manzil Zaheer · Yinlam Chow · Amr Ahmed · Craig Boutilier -
2020 Poster: Big Bird: Transformers for Longer Sequences »
Manzil Zaheer · Guru Guruganesh · Kumar Avinava Dubey · Joshua Ainslie · Chris Alberti · Santiago Ontanon · Philip Pham · Anirudh Ravula · Qifan Wang · Li Yang · Amr Ahmed -
2018 Poster: Learning Pipelines with Limited Data and Domain Knowledge: A Study in Parsing Physics Problems »
Mrinmaya Sachan · Kumar Avinava Dubey · Tom Mitchell · Dan Roth · Eric Xing -
2014 Poster: Dependent nonparametric trees for dynamic hierarchical clustering »
Kumar Avinava Dubey · Qirong Ho · Sinead Williamson · Eric Xing