| Project/Area Number |
23KJ0616
|
|---|
| Research Category |
Grant-in-Aid for JSPS Fellows
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 国内 |
|---|
| Review Section |
Basic Section 60080:Database-related
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
CHEN Jinyu 東京大学, 大学院情報学環, 特別研究員(PD)
|
|---|
| Project Period (FY) |
2023-04-25 – 2025-03-31
|
| Project Status |
Granted (Fiscal Year 2023)
|
| Budget Amount *help |
¥1,800,000 (Direct Cost: ¥1,800,000)
Fiscal Year 2024: ¥900,000 (Direct Cost: ¥900,000)
Fiscal Year 2023: ¥900,000 (Direct Cost: ¥900,000)
|
|---|
| Keywords | Deep learning / Artificial Intelligence |
|---|
| Outline of Research at the Start |
In this research, I plan to develop two connected methods that separately address the problem of 1. How to find out the significant input data for prediction result in heterogeneous input dataset that requires lots of efforts for collection. 2. How to detect false input data and protect the result.
|
| Outline of Annual Research Achievements |
The research is going smoothly. My research proposal is divided into two parts. In the first year, I’ve completed the first part, which is designated to a game theory-based input scorer. In deep learning, the input variables are usually coupled and interacted to give the precise prediction. The scorer groups the input variables into different groups and compute the Shapley value of each pair of groups. In this year, for my personal use, I used the developed scorer to help me develop two deep learning methods for the time series prediction and successfully made them published in international journal in the filed of Artificial Intelligence.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
Based on the research proposal, two works have been published separetely in two international journals on artificial intelligence and computer science. The first is the full titled as “Mutual adaptation: learning from prototype for time series prediction”, published in “IEEE Transactions on Artificial Intelligence”; the second is named "MobCovid: Confirmed Cases Dynamics Driven Time Series Prediction of Crowd in Urban Hotspot" published in "IEEE Transactions on Neural Networks and Learning Systems" .
|
| Strategy for Future Research Activity |
I plan to finish the second part of my research proposal and simultanesouly, publish more papers in this fiscal year.For the time being, one paper is under preparation and there are some experiments required, which will request more computation resources. This year will be needing more computation resources like GPUs. So, I plan to purchase some GPU computation resource and build up a computation PC to make the research go smoothly.
|
