Aaron Thomas Roth

Featured in: Favicon

arxiv.org

scientificamerican.com Favicon

brookings.edu Favicon

acm.org

motortrend.com Favicon

shine.cn

strategy-business.com Favicon

hstoday.us Favicon

amazon.science Favicon

fivebooks.com + 4 more

Articles

[2409.14513] Order of Magnitude Speedups for LLM Membership Inference

Sep 22, 2024 | arxiv.org | Rongting Zhang |Aaron Thomas Roth
Bias bounties

Jan 25, 2024 | amazon.science | Michael Kearns |Aaron Thomas Roth |Xiangkun Hu |Dongyu Ru

Project DescriptionThis is a test framework for the bias bounties project. Getting Started as a Bounty HunterIf you are interacting with this codebase as a "bounty hunter", you'll need to have a way to run Jupyter notebooks. The easiest way to do this is to download Anaconda, which will also manage all of your python packages for you. See here for installation instructions: https://docs.anaconda.com/anaconda/install/index.html.
Improved differentially private regression via gradient boosting

Jan 19, 2024 | amazon.science | Shuai Tang |Sergul Aydore |Michael Kearns |Aaron Thomas Roth

We revisit the problem of differentially private squared error linear regression. We observe that existing state- of-the-art methods are sensitive to the choice of hyperparameters — including the “clipping threshold” that cannot be set optimally in a data-independent way. We give a new algorithm for private linear regression based on gradient boosting.
Responsible AI in the wild: Lessons learned at AWS

Nov 16, 2023 | amazon.science | Michael Kearns |Aaron Thomas Roth |Changyou Chen |Daniel Zhang

When we first joined AWS AI/ML as Amazon Scholars over three years ago, we had already been doing scientific research in the area now known as responsible AI for a while. We had authored a number of papers proposing mathematical definitions of fairness and machine learning (ML) training algorithms enforcing them, as well as methods for ensuring strong notions of privacy in trained models.
Aaron Roth

Nov 16, 2023 | amazon.science | Michael Kearns |Aaron Thomas Roth |Sergul Aydore |Shuai Tang

Aaron Roth is the Henry Salvatori Professor of Computer and Cognitive Science at the University of Pennsylvania and an Amazon Scholar. His research focuses on the algorithmic foundations of data privacy, algorithmic fairness, game theory, learning theory, and machine learning. Together with Cynthia Dwork, he is the author of the book The Algorithmic Foundations of Differential Privacy; together with Michael Kearns, he is the author of The Ethical Algorithm. Read more
[2306.15083] Balanced Filtering via Non-Disclosive Proxies

Jun 26, 2023 | arxiv.org | Michael Kearns |Aaron Thomas Roth |Emily Diana

[Submitted on 26 Jun 2023 (v1), last revised 5 Jul 2023 (this version, v2)] Title:Balanced Filtering via Non-Disclosive Proxies Download a PDF of the paper titled Balanced Filtering via Non-Disclosive Proxies, by Siqi Deng and 3 other authors Download PDF Abstract: We study the problem of non-disclosively collecting a sample of data that isbalanced with respect to sensitive groups when group membership is unavailableor prohibited from use at collection time.
PERSPECTIVE: A Call to Action to Resource Our Nation’s Security

Mar 27, 2023 | hstoday.us | Aaron Thomas Roth

Decreased funding is not the only hurdle facing DHS, but highly politicized debate over DHS mission priorities, ongoing legislative challenges, and the continued reliance on continuing resolutions. President Biden released his annual budget request on March 9. Given Republican House control, final appropriations are likely to look very different from this request. However, the president’s budget typically serves as a strong indicator of the administration’s priorities.