On the learning coefficients in learning theory
- Hold Date
- 2021-12-21 17:30~2021-12-21 18:30
- Place
- Zoom
- Object person
-
- Speaker
-
Miki Aoyagi (College of Science and Technology, Nihon University)
The IMI Colloquium in December 2021 | Date : Wednesday, 21 December 2021 17:30-18:30 | Place : Llive-streaming by Zoom | | Speaker : Miki Aoyagi (College of Science and Technology, Nihon University) | | Title : | | On the learning coefficients in learning theory | | Abstract : In recent studies, image or speech recognition, probabilistic inference, genetic analysis, time series prediction, and data mining, etc. In real big data have been analyzed by learning systems. They are very complicated and not usually generated by a simple normal distribution, as they are influenced by many factors. Hierarchical learning models such as layered neural networks, the normal mixture model, and reduced rank regression may be known as effective learning models. They, however, likewise have complicated, i.e., singular structures, which cannot be analyzed using the classic theories of regular statistical models. The theoretical study has therefore been started to construct a mathematical foundation for singular statistical models. In this talk, we consider learning coefficients in learning theory, which serve to measure the main term of learning efficiency in singular learning models. These coefficients have an important role in information criteria and are mathematically equal to the log canonical thresholds which are obtained by using complete de-singularization. | | | | |
*** Notes *** |
| |
• | If you have not downloaded the Zoom application yet, please download it (https://zoom.us) in advance. |
• | The account name should be "Full name" in case of the faculty members, and "Student ID number(Last name)" (e.g.:1SC○○○○○A(Last name) in case of students. |
• | Please understand the host will mute the audience's sound and camera. |
• | When you have a question, please use "Raise hand function" or "Chat function". Then, host will unmute. |
• | Login becomes possible half an hour before. |
| |
If you have any questions, please contact Tagami-sensei (E-mail: tagami(at)imi.kyushu-u.ac.jp). |