Presentation 2023-01-19
[Short Paper] A Study of Joint Control Method of Wireless LAN and Machine Learning Settings for Communication-efficient Split Computing
Kojin Yorita, Takayuki Nishio, Daiki Yoda, Toshihisa Nabetani,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Split computing (SC) enables machine learning (ML) inference with a deep neural network on resource-constrained devices. However, a narrow-bandwidth and lossy wireless network can become a bottleneck, thereby increasing communication latency in SC due to retransmission and the use of lower physical transmission rate. To solve this trade-off between communication latency and inference accuracy, this paper studies a joint control of the wireless communication parameters (e.g., transmission rate, retransmission limit) and ML model to reduce communication latency and improve inference accuracy. The proposed method focuses on the packet-loss tolerance of ML inference and uses unreliable (i.e., high packet-loss rate) but low-latency communication protocol. To achieve a well-balanced trade-off between communication latency and inference accuracy, the proposed method jointly controls wireless communication parameters affecting the reliability and communication latency and ML model architecture affecting inference accuracy and packet-loss reliance, based on multi-armed bandit (MAB) algorithm, namely upper confidence bound (UCB) algorithm. The results of ns3-based computer simulations show that the proposed method reduces communication latency and improves inference accuracy.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Split computing / Joint control / Machine learning / Wireless LAN / Reinforcement Learning
Paper # SeMI2022-77
Date of Issue 2023-01-12 (SeMI)

Conference Information
Committee SeMI
Conference Date 2023/1/19(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Naruto grand hotel
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair Koji Yamamoto(Kyoto Univ.)
Vice Chair Kazuya Monden(Hitachi) / Yasunori Owada(NICT) / Shunsuke Saruwatari(Osaka Univ.)
Secretary Kazuya Monden(NTT DOCOMO) / Yasunori Owada(Tokyo Univ. of Agri. and Tech.) / Shunsuke Saruwatari(Osaka Univ.)
Assistant Yuki Matsuda(NAIST) / Akihito Taya(Aoyama Gakuin Univ.) / Takeshi Hirai(Osaka Univ.)

Paper Information
Registration To Technical Committee on Sensor Network and Mobile Intelligence
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) [Short Paper] A Study of Joint Control Method of Wireless LAN and Machine Learning Settings for Communication-efficient Split Computing
Sub Title (in English)
Keyword(1) Split computing
Keyword(2) Joint control
Keyword(3) Machine learning
Keyword(4) Wireless LAN
Keyword(5) Reinforcement Learning
1st Author's Name Kojin Yorita
1st Author's Affiliation Tokyo Institute of Technology(Tokyo Tech)
2nd Author's Name Takayuki Nishio
2nd Author's Affiliation Tokyo Institute of Technology(Tokyo Tech)
3rd Author's Name Daiki Yoda
3rd Author's Affiliation Toshiba(Toshiba)
4th Author's Name Toshihisa Nabetani
4th Author's Affiliation Toshiba(Toshiba)
Date 2023-01-19
Paper # SeMI2022-77
Volume (vol) vol.122
Number (no) SeMI-341
Page pp.pp.28-29(SeMI),
#Pages 2
Date of Issue 2023-01-12 (SeMI)