【115/4/16】15:20-16:10 薛名成 教授(國立陽明交通大學應用數學系）
發佈日期 :
2026-04-02
| Colloquium | |
|---|---|
|
|
|
| Time | 2026-4-16 15:20-16:10 |
| Venue | 數學館1樓31106教室 |
| Speaker | 薛名成 教授(國立陽明交通大學應用數學系） |
| Title | Kolmogorov-Arnold Theorem and Its Connection to KAN Neural Networks |
| Abstract | The Kolmogorov-Arnold Representation Theorem (KAT) is a foundational result in mathematical analysis that states any multivariate continuous function can be represented as a superposition of continuous univariate functions and a finite number of additions. This theorem has played a crucial role in the development of function approximation theories and has recently gained renewed interest as the theoretical backbone of Kolmogorov-Arnold Networks (KANs). In this talk, I will review the historical development and refinements of the Kolmogorov-Arnold Theorem and the approximate capability of the KANs by using polynomials is presented. |
|
|
|
