The 3rd KMGS will be held on March 17th, Thursday, via Zoom and Gather Town.
We invite Junghyun Lee who graduated from Dept. of Mathematical Sciences, KAIST and is now a graduate student in Kim Jaechul Graduate School of AI, KAIST.
In this seminar, we have one talk which will last for 40 minutes with 10 minutes discussion session, after the talk.
The abstract is as follows.
Slot (PM 12:00~12:40)
[Speaker] Junghyun Lee (이정현) from Kim Jaechul Graduate School of AI, KAIST, supervised by Prof. Se-Young Yun (윤세영 교수님)
[Title] Fast and efficient MMD-based fair PCA via optimization over Stiefel manifold
[Discipline] Machine Learning, Optimization
[Abstract]
This paper defines fair principal component analysis (PCA) as minimizing the maximum mean discrepancy (MMD) between dimensionality-reduced conditional distributions of different protected classes. The incorporation of MMD naturally leads to an exact and tractable mathematical formulation of fairness with good statistical properties. We formulate the problem of fair PCA subject to MMD constraints as a non-convex optimization over the Stiefel manifold and solve it using the Riemannian Exact Penalty Method with Smoothing (REPMS; Liu and Boumal, 2019). Importantly, we provide local optimality guarantees and explicitly show the theoretical effect of each hyperparameter in practical settings, extending previous results. Experimental comparisons based on synthetic and UCI datasets show that our approach outperforms prior work in explained variance, fairness, and runtime.
This paper is accepted to the 36th AAAI Conference on Artificial Intelligence (AAAI 2022).
[Language] Korean (English if it is requested)
[Zoom link]
https://kaist.zoom.us/j/2655728482?pwd=OXpJeFdDcWliSG51WUp0N1Nad2JHdz09
ID: 265 572 8482
Password: 2AHRKr
[Gather Town link]
https://gather.town/app/ffr2PVibAWRIyXWO/kaistmath