HIAS Brown Bag Seminar #7
Date: Thursday 19, September, 2024 12:40-13:40
Title: “Domain adaptation for extra features using optimal transport”
Speaker: Toshimitsu Aritake (Assistant Professor, HIAS)
Abstract: “In conventional supervised machine learning, the assumption that the distributions of training and test data are identical underpins the effectiveness of models trained on labeled training data for unseen test data. However, in real-world applications, this assumption often fails, leading to suboptimal performance during the deployment phase. To mitigate this issue, domain adaptation frameworks are employed to bridge the domain gap by minimizing distribution discrepancies between training and test datasets.
This talk will explore domain adaptation methods grounded in optimal transport theory, with a focus on scenarios where additional features are available exclusively during the test phase, while a common set of features is shared across both the training and test phases. The discussion will introduce a method based on bi-directional optimal transport and a novel approach leveraging fused Gromov–Wasserstein optimal transport, which simultaneously addresses both standard and Gromov–Wasserstein optimal transport problems.”
(*) No. 5 building in this campus map
Language: English
<If space allows, walk-in participants will be accepted on the day of the event.>
*Bring your own lunch.
<Coffee and snacks will be served.>
Click HERE to register!
