Integration of machine learning and mathematical modeling, and deepening of its theory Ⅲ
| Reference No. | 2025a043 |
|---|---|
| Type/Category | Grant for General Research-Workshop（Ⅱ） |
| Title of Research Project | Integration of machine learning and mathematical modeling, and deepening of its theory Ⅲ |
| Principal Investigator | Takiko Sasaki（Department of Mathematical Engineering, Faculty of Engineering, Musashino University・Associate Professor） |
| Research Period |
October 11,2024. -
October 13,2024. |
| Keyword(s) of Research Fields | Machine learning, natural language processing, mathematical modeling |
| Abstract for Research Report |
(Purpose) This workshop is positioned as a continuation of the research symposium, "Integration of machine learning and mathematical modeling, and deepening of its theory II" held in October 2024. Focusing on two research themes, ① "Proposal of Predictive Control Methods Combining Machine Learning and Mathematical Models" and ② "Mathematics for Natural Language Processing", we aim to discuss both theoretical and practical aspects, identify new challenges, and build a cross-disciplinary research collaboration framework. ① : In recent years, the functions and performance required of systems have become increasingly sophisticated and complex, requiring control technology that can withstand advanced and complex constraints. In this research, mathematicians and researchers specializing in machine learning discuss and work on “model prediction control”, a control method that performs optimization while predicting future actions at each time by using an appropriate mathematical model to predict actions and determining actions so that the function which expresses the desirability of the outcome is the largest. ② : Although demand for natural language processing is increasing due to its usefulness, and future development is expected, it has specific difficulties, such as ambiguity in interpreting the meaning of words. For example, the distributed representation of words is essentially a one-to-one relationship, with one vector corresponding to one word, and requires refinement from a mathematical perspective, such as how to handle word polysemy and how to evaluate the distributed representation obtained through learning. In this study, mathematicians and researchers specializing in natural language processing will discuss and explore the possibility of building robust natural language processing systems. (Expected Outcomes): By bringing together researchers and graduate students from various fields of mathematical engineering, such as topology and numerical analysis, and discussing the contents of their presentations, it is expected to discover potential applications and implementations of machine learning techniques in various fields, create new machine learning methods, and initiate joint research with companies. |
| Organizing Committee Members (Workshop) Participants (Short-term Joint Usage) |
Takashi Tsuboi（Tohoku Forum for Creativity, Tohoku University・Professor） Tetsuji Tokihiro（Department of Mathematical Engineering, Faculty of Engineering, Musashino University・Professor） Osamu Saeki（Institute of Mathematics for Industry, Kyushu University・Professor） Hiroyuki Ochiai（Institute of Mathematics for Industry, Kyushu University・Professor） Kohei Higashi（Department of Mathematical Engineering, Faculty of Engineering, Musashino University・Assistant Professor） |