Presentation | 2024-02-29 Model Shifting Method in Federated Learning Using Distillation Hiromichi Yajima, Shota Ono, Takumi Miyoshi, Taku Yamazaki, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Due to the drastic increase in the data for machine learning, distributed machine learning such as federated learning has been attracting attention to avoid the intensive load on a server. Since the process of machine learning requires a huge amount of computation, it is difficult to perform federated learning process on low-performance client devices. To solve this problem, federated learning with downsizing machine learning model by distillation has been proposed. This method can reduce the processing cost and learning time on the clients but degrades the achieved accuracy of machine learning model. This paper proposes a method that improves the accuracy of machine learning model in a short time with downsized machine learning model in the early stage of learning and then switches to general federated learning process. From the experimental results, the proposed method can rapidly improve the accuracy of machine learning model while counteracting the degradation of the final accuracy achieved. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Machine learning / Federated learning / Distillation / Machine learning model |
Paper # | NS2023-186 |
Date of Issue | 2024-02-22 (NS) |
Conference Information | |
Committee | NS / IN |
---|---|
Conference Date | 2024/2/29(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Okinawa Convention Center |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | General |
Chair | Tetsuya Oishi(NTT) / Kunio Hato(NTT) |
Vice Chair | Takumi Miyoshi(Shibaura Inst. of Tech.) / Tsutomu Murase(Nagoya Univ.) |
Secretary | Takumi Miyoshi(NTT) / Tsutomu Murase(Kogakuin Univ.) |
Assistant | Hiroshi Yamamoto(NTT) |
Paper Information | |
Registration To | Technical Committee on Network Systems / Technical Committee on Information Networks |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Model Shifting Method in Federated Learning Using Distillation |
Sub Title (in English) | |
Keyword(1) | Machine learning |
Keyword(2) | Federated learning |
Keyword(3) | Distillation |
Keyword(4) | Machine learning model |
1st Author's Name | Hiromichi Yajima |
1st Author's Affiliation | Shibaura Institute of Technology(SIT) |
2nd Author's Name | Shota Ono |
2nd Author's Affiliation | The University of Tokyo(The Univ. of Tokyo) |
3rd Author's Name | Takumi Miyoshi |
3rd Author's Affiliation | Shibaura Institute of Technology(SIT) |
4th Author's Name | Taku Yamazaki |
4th Author's Affiliation | Shibaura Institute of Technology(SIT) |
Date | 2024-02-29 |
Paper # | NS2023-186 |
Volume (vol) | vol.123 |
Number (no) | NS-397 |
Page | pp.pp.86-89(NS), |
#Pages | 4 |
Date of Issue | 2024-02-22 (NS) |