Working at Mila

Being surrounded by hundreds of AI researchers at Mila creates an exceptional environment to develop ideas and make new friends!

What I like working on

I like working on responsible AI-related topics, such as developing novel and efficient social bias mitigation methods and studying harms like hate speech and hallucination. Given the limitations of bias metrics, I am also interested in introducing better fairness assessment techniques.

Publications

Do Biased Models Have Biased Thoughts?
[arXiv] [slides] [5-minute video] [1-hour video]

Should we attend more or less? modulating attention for fairness

A. Zayed, Goncalo Mordido, Samira Shabanian, Sarath Chandar
1st Conference on Language Modeling (COLM 2024)
TL;DR: We show that bias can be mitigated by focusing on a wider/narrower context.
[arXiv] [blog]

Why Don’t Prompt-Based Fairness Metrics Correlate?

A. Zayed, Goncalo Mordido, Ioana Baldini, Sarath Chandar
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024)
TL;DR: We explain why fairness metrics don't correlate and propose CAIRO to make them correlate.
[arXiv] [slides] [video] [longer video] [blog]

Fairness-Aware Structured Pruning in Transformers

A. Zayed, Goncalo Mordido, Samira Shabanian, Ioana Baldini, Sarath Chandar
The thirty-eighth Association for the Advancement of Artificial Intelligence (AAAI 2024)
TL;DR: We show that certain attention heads are responsible for bias and pruning them improves fairness.
[arXiv] [slides] [video] [blog]

Deep Learning on a Healthy Data Diet: Finding Important Examples for Fairness

A. Zayed, Prasanna Parthasarathi, Goncalo Mordido, Hamid Palangi, Samira Shabanian, Sarath Chandar
The thirty-seventh Association for the Advancement of Artificial Intelligence (AAAI 2023)
TL;DR: We quantify the contribution of each training example to fairness, and choose the best training examples to reduce bias.
[arXiv] [slides] [video] [blog]

Ultrasound Elastography using Machine Learning
A. Zayed
Master's Thesis, 2020
[Link]

Fast Strain Estimation and Frame Selection in Ultrasound Elastography using Machine Learning
A. Zayed and H. Rivaz
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control (TUFFC), 2020
[Link] [code]

Automatic Frame Selection using CNN in Ultrasound Elastography
A. Zayed, G. Cloutier and H. Rivaz
Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 2020
[arXiv]

Rapid Quantification of White Matter Disconnection in the Human Brain
A. Zayed, Y. Iturria-Medina, A. Villringer, B. Sehm and C. Steele
Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 2020
[arXiv]

Automatic Frame Selection Using MLP Neural Network in Ultrasound Elastography
A. Zayed and H. Rivaz
International Conference on Image Analysis and Recognition (ICIAR), 2019
[arXiv]

Fast approximate time-delay estimation in ultrasound elastography using principal component analysis
A. Zayed and H. Rivaz
Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 2019
[arXiv]