Data-efficient Low-Rank Tensor Completion via Integer Optimization
김수현
27동 220호
0
104
12.05 17:36
| 구분 | 응용수학 |
|---|---|
| 일정 | 2026-01-06(화) 16:00~17:30 |
| 세미나실 | 27동 220호 |
| 강연자 | Chen Chen (The Ohio State University) |
| 담당교수 | 이다빈 |
| 기타 |
In machine learning, tensors can be viewed as a flexible way to structure data. Tensors generalize matrices and can, in principle, be leveraged to deliver more accurate predictions. However, this more powerful modeling paradigm comes with substantial computational cost: for example, although matrix rank can be calculated quickly, determining the rank of even a 3-tensor is an NP-hard computational challenge. This talk discusses our latest results in developing practical, data-efficient algorithms for tensor completion. Tensor completion is a problem of recovering an underlying tensor from partial, noisy measurements. This paradigm has seen widespread application over the last decade in recommendation engines, sensor fusion, network analysis, image processing, etc.
