| Presentation | 2022-12-13 A Study On the Impact of Network Topology on the Efficiency of Distributed Online Kernel Learning Koki Takamori, Taichi Emi, Han Nay Aung, Keita Goto, Hiroyuki Ohsaki, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Distributed learning, which estimates the parameters of a nonlinear model from nonlinear data observed at each distributed node in a network without aggregating the data at a single location, has been attracting attention. In particular, by approximating the kernel function using random Fourier features distributed and online nonlinear learning. Bouboulis et al. have proposed a distributed online kernel-based learning algorithm RFF-DOKL (Random Fourier Features Distributed Online Kernel-based Learning) using random Fourier features. On the other hand, the effect of network topology on the efficiency of RFF-DOKL has not been fully clarified. In this paper, we experimentally investigate the effect of network topology on the efficiency of RFF-DOKL, distributed online kernel-based learning algorithm. Specifically, we experimentally analyze the relationship between total traffic and model accuracy when distributed online learning of nonlinear functions is performed using RFF-DOKL with Gaussian kernels on four different network topologies (series, ring, star, and mesh) with the same number of nodes. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | RFF (Random Fourier Features) / RFF-DOKL (Random Fourier Features Distributed Online Kernel-based Learning) / Kernel-based Learning / Distributed Learning / Network Topology |
| Paper # | IA2022-58 |
| Date of Issue | 2022-12-05 (IA) |
| Conference Information | |
| Committee | IN / IA |
|---|---|
| Conference Date | 2022/12/12(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Higashi-Senda campus, Hiroshima Univ. |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Performance Analysis and Simulation, Robustness, Traffic and Throughput Measurement, Quality of Service (QoS) Control, Congestion Control, Overlay Network/P2P, IPv6, Multicast, Routing, DDoS, etc. |
| Chair | Kunio Hato(Internet Multifeed) / Tomoki Yoshihisa(Osaka Univ.) |
| Vice Chair | Tsutomu Murase(Nagoya Univ.) / Yusuke Sakumoto(Kwansei Gakuin Univ.) / Yuichiro Hei(KDDI Research) / Hiroshi Yamamoto(Ritsumeikan Univ.) |
| Secretary | Tsutomu Murase(KDDI Research) / Yusuke Sakumoto(Nagaoka Univ. of Tech.) / Yuichiro Hei(NTT) / Hiroshi Yamamoto(NTT) |
| Assistant | / Daisuke Kotani(Kyoto Univ.) / Ryo Nakamura(Fukuoka Univ.) / Ryo Nakamura(Univ. of Tokyo) |
| Paper Information | |
| Registration To | Technical Committee on Information Networks / Technical Committee on Internet Architecture |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Study On the Impact of Network Topology on the Efficiency of Distributed Online Kernel Learning |
| Sub Title (in English) | |
| Keyword(1) | RFF (Random Fourier Features) |
| Keyword(2) | RFF-DOKL (Random Fourier Features Distributed Online Kernel-based Learning) |
| Keyword(3) | Kernel-based Learning |
| Keyword(4) | Distributed Learning |
| Keyword(5) | Network Topology |
| 1st Author's Name | Koki Takamori |
| 1st Author's Affiliation | Kwaisei Gakuin University(Kwansei Univ.) |
| 2nd Author's Name | Taichi Emi |
| 2nd Author's Affiliation | Kwaisei Gakuin University(Kwansei Univ.) |
| 3rd Author's Name | Han Nay Aung |
| 3rd Author's Affiliation | Kwaisei Gakuin University(Kwansei Univ.) |
| 4th Author's Name | Keita Goto |
| 4th Author's Affiliation | Kwaisei Gakuin University(Kwansei Univ.) |
| 5th Author's Name | Hiroyuki Ohsaki |
| 5th Author's Affiliation | Kwaisei Gakuin University(Kwansei Univ.) |
| Date | 2022-12-13 |
| Paper # | IA2022-58 |
| Volume (vol) | vol.122 |
| Number (no) | IA-306 |
| Page | pp.pp.56-59(IA), |
| #Pages | 4 |
| Date of Issue | 2022-12-05 (IA) |