2021 Fiscal Year Annual Research Report
Life-Long Deep Learning using Bayesian Principles
| Project/Area Number |
20H04247
|
|---|
| Research Institution | Institute of Physical and Chemical Research |
|---|
Principal Investigator |
Khan Emtiyaz 国立研究開発法人理化学研究所, 革新知能統合研究センター, チームリーダー (30858022)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
Alquier Pierre 国立研究開発法人理化学研究所, 革新知能統合研究センター, 研究員 (10865645)
横田 理央 東京工業大学, 学術国際情報センター, 教授 (20760573)
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | continual learning / adaptation |
|---|
| Outline of Annual Research Achievements |
This fiscal year we continued working on continual learning. Building up on our previous paper, we focused on developing fundamental principles of adaptation. One research paper was published at NeurIPS 2021 as a poster presentation.
Khan, M. E. and Swaroop, S. "Knowledge-adaptation priors" Advances in Neural Information Processing Systems 34, pp. 19757-19770, (2021).
This work proposes a new principle of adaptation which most machine-learning models are expected to follow. The paper provides a solid foundation for our previous work. We hope to build on this work to scale our algorithm to ImageNet level.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
We build upon our previous work and proposed a new solid foundational principle for it.
|
| Strategy for Future Research Activity |
We will continue towards our main goal to run a continual learning algorithm on ImageNet.
|
-
[Journal Article] Knowledge-Adaptation Priors2021
Author(s)
Khan, Mohammad Emtiyaz E and Swaroop, Siddharth
-
Journal Title
Advances in Neural Information Processing Systems
Volume: 34
Pages: 19757--19770
Peer Reviewed / Open Access / Int'l Joint Research
-
-
-
