About Me

Yoshitomo Matsubara

Ph.D. candidate
Department of Computer Science
Donald Bren School of Information & Computer Sciences
University of California, Irvine
Latest News

November 25, 2020
My paper was accepted to ICPR 2020 Workshop on Reproducible Research in Pattern Recognition.
November 20, 2020
Our paper presented at EMNLP 2020 NLP COVID-19 Workshop (Part 2) received the Best Paper Award.
October 30, 2020
I am volunterring for the first fully virtual NeurIPS 2020.
October 17, 2020
Our papper was accepted to IEEE Access.
October 1, 2020
Our papper was accepted to EMNLP 2020 NLP COVID-19 Workshop (Part 2).
September 28, 2020
This quarter, I am working as a graduate student researcher and for Amazon.com Services, Inc. as an applied scientist intern.
August 16, 2020
Our papper was accepted to GLOBECOM 2020.
August 2, 2020
Our papper was accepted to ICPR 2020.
July 28, 2020
Our papper was accepted to MobiCom 2020 Workshop on Embedded and Mobile Deep Learning.
July 27, 2020
Our papper was accepted to JCDL 2020 Workshop on Mining Scientific Publications.
July 1, 2020
I am volunterring for the first fully virtual ICML 2020.
June 22, 2020
This quarter, I am working for Amazon.com Services, Inc. as an applied scientist intern.
May 22, 2020
Our invention of Flick Authentication is registered as JP Patent No. 6707271.
April 24, 2020
My work with Thuy Vu and Alessandro Moschitti at Amazon.com Service, Inc. was accepted as a short paper for SIGIR 2020.
April 8, 2020
I am volunterring for the first fully virtual ICLR 2020.
March 25, 2020
This quarter, I am working as a graduate student researcher.
March 6, 2020
We, UCI Table Tennis Club Varsity team, are invited to the 2020 iSET College Table Tennis National Championships.
February 29, 2020
We, UCI Table Tennis Club Varsity team, placed 4th in NCTTA West Regional Championships.
February 1, 2020
We, UCI Table Tennis Club Varsity team, won the Socal East Division.
January 2, 2020
This quarter, I am working as a graduate student researcher.
October 21, 2019
I presented our work at MobiCom 2019 Workshop on Hot Topics in Video Analytics and Intelligent Edges.
September 23, 2019
This quarter, I am working as a graduate student researcher.
September 23, 2019
In 2019-2020 season, I serve as a vice president of UCI Table Tennis Club.
September 23, 2019
Our papper was accepted to MobiCom 2019 Workshop on Hot Topics in Video Analytics and Intelligent Edges.
September 17, 2019
I was invited to talk at Robot Learning Group in Osaka University, Japan.
June 24, 2019
This quarter, I am working for Amazon.com Services, Inc. as an applied scientist intern.
June 13, 2019
I passed my Ph.D. qualifying exam, and successfully advanced to candidacy.
June 5, 2019
Our team, Zotbot, was selected as a finalist (10 of 375 teams) for the Alexa Prize Socialbot Grand Challenge 3.
May 23, 2019
We, UCI Table Tennis Club Varsity team, was featured on UCI News "UCI table tennis team competes for first time in national championships, takes fourth place".
April 14, 2019
We, UCI Table Tennis Club Varsity team, placed 4th in iSET College Table Tennis National Championships.
March 27, 2019
This quarter, I am assisting CS 273P: "Machine Learning & Data Mining".
March 27, 2019
We, UCI Table Tennis Club Varsity team, are featured on NCTTA, "College Teams Embrace First Trip to Table Tennis National Championships".
March 9, 2019
We, UCI Table Tennis Club Varsity team, won the NCTTA West Regional Championships, and advanced to the 2019 iSET College Table Tennis National Championships.
February 9, 2019
We, UCI Table Tennis Club Varsity team, won the Socal East Division.
January 7, 2019
This quarter, I am assisting CS 272: "Statistical NLP".
November 17, 2018
We, UCI Table Tennis Club varsity team, won the Socal East & West Division.
September 24, 2018
This quarter, I am assisting CS 273A: "Machine Learning".
June 25, 2018
This quarter, I am working for Slice Technologies, Inc. as a machine learning engineeering intern.
April 3, 2018
This quarter, I am assisting BANA 290: "Advanced Data Analytics for Natural Language Processing".
January 3, 2018
This quarter, I am assisting CS 273A: "Machine Learning".
September 28, 2017
This quarter, I am assisting ICS 32A: "Python Programming with Libraries (Accelerated)".
July 20, 2017
I was invited to speak on "Seminar of Study Abroad for Pursuing a Degree" at Global Terrace, National Institute of Technology, Akashi College.
June 26, 2017
I started to work for Data & Science Solutions Group, Yahoo Japan Corporation as a data scientist (contract employee).
April 3, 2017
This quarter, I am assisting CS 122B: "Projects in Databases and Web Applications".
January 9, 2017
This quarter, I am assisting ICS 45C: "Programming in C/C++ as a Second Language".
September 22, 2016
This quarter, I am assisting ICS 45J: "Programming in Java as a Second Language".
September 19, 2016
Our paper on flick authentication was accepted to ICMLA 2016.
September 19, 2016
I enrolled in Donald Bren School of Information & Computer Sciences (Computer Science Ph.D. program) at the University of California, Irvine.
June 9, 2016
I was invited to speak on "Study Abroad to Pursue a Graduate Degree from "Zero" -Introduction and Preparation-" at Global Terrace, National Institute of Technology, Akashi College.
April 1, 2016
I joined the Information Science Research Center for Social Applications in the Graduate School of Applied Informatics at the University of Hyogo as a project researcher.
March 22, 2016
I graduated from Graduate School of Applied Informatics (High Confidence Informatics Course ) at the University of Hyogo and obtained a Master's degree in Applied Informatics.
January 25, 2016
I started to work for Data & Science Solutions Group, Yahoo Japan Corporation as a data scientist (contract employee).
November 30, 2015
I worked for Yahoo Japan Corporation as a special intern (data science).
November 12, 2015
I did a poster presentation at The Fifth Symposium on Biometrics, Recognition and Authentication(SBRA2015).
October 25, 2015
Our technical report was published in IEICE Technical Report.
July 14, 2015
Our paper on keystroke dynamics was published in Journal of Information Security.
June 12, 2015
I obtained the Grant Scholarship for Japanese Students from JGC-S Scholarship Foundation.
April 16, 2015
Our paper on keystroke dynamics was published in Journal of Information Security.
February 27, 2015
I worked for Recruit Holdings Co., Ltd. as a outsourced worker.
September 19, 2014
I took part in Recruit Holdings Summer Internship 2014 Web Innovation Course Data Analysis Internship, and won the victory in the team competition.
April 4, 2014
I enrolled in Graduate School of Applied Informatics (High Confidence Informatics Course ) at the University of Hyogo.
March 20, 2014
I graduated from Advanced Course of Mechanical and Electronic System Engineering, Akashi National College of Technology and obtained a Bachelor's degree in Engineering.
March 20, 2014
I received President's Award in INCT's Student Commendation from the Institute of National Colleges of Technology, Japan.
March 1, 2014
I opened my home page.

Projects
NLP COVID-19
2020 - Present
Knowledge Distillation for Split Computing
2018 - Present
Alexa Prize SGD3
2019 - 2020
Mining Scientific Papers
2016 - 2018
Flick Authentication
2014 - 2016
Safe Driving Support
2013 - 2014
TESOL
2013 - 2014
Keystroke Dynamics
2010 - 2014

Knowledge Distillation for Split Computing



Offloading the execution of complex Deep Neural Networks (DNNs) models to compute-capable devices at the network edge, that is, edge servers, can significantly reduce capture-to-output delay. However, the communication link between the mobile devices and edge servers can become the bottleneck when channel conditions are poor. We propose a framework to split DNNs for image processing and minimize capture-to-output delay in a wide range of network conditions and computing parameters. The core idea is to split the DNN models into head and tail models, where the two sections are deployed at the mobile device and edge server, respectively. Following the trends of mobile and edge computing for DNN models, an intermediate option, split computing, has been attracting attentions from the research community. Previous studies empirically showed that while mobile and edge computing often would be the best options in terms of total inference time, there are some scenarios where split computing methods can achieve shorter inference time.


University of California, Irvine [4/2018 - present]
  • Graduate Student Researcher
  • Advisor: Prof. Marco Levorato (Department of Computer Science, University of California, Irvine)
    Advisor: Prof. Sameer Singh (Department of Computer Science, University of California, Irvine)

  • Studying about learning representations of bottlenecks injected to DNN models for split computing, that is an intermediate option between mobile and edge computing for achieving shorter inference time in challenged network conditions, while preserving the inference performance of the original models (without bottlenecks).


Papers

  • Peer-reviewed Conference/Workshop Papers
    • D. Callegaro, Y. Matsubara and M. Levorato, "Optimal Task Allocation for Time-Varying Edge Computing Systems with Split DNNs", Proceedings of the 2020 IEEE Global Communications Conference (GLOBECOM '20), (2020)
    • Y. Matsubara and M. Levorato, "Neural Compression and Filtering for Edge-assisted Real-time Object Detection in Challenged Networks", Proceedings of the 25th International Conference on Pattern Recognition (ICPR '20), (2021)
      [Preprint] [Code]
    • Y. Matsubara and M. Levorato, "Split Computing for Complex Object Detectors: Challenges and Preliminary Results", Proceedings of the 4th International Workshop on Embedded and Mobile Deep Learning (EMDL '20), pp. 7-12 (2020)
      [PDF (Open Access)] [Preprint] [Code]
    • Y. Matsubara, S. Baidya, D. Callegaro, M. Levorato and S. Singh, "Distilled Split Deep Neural Networks for Edge-Assisted Real-Time Systems", Proceedings of the 2019 Workshop on Hot Topics in Video Analytics and Intelligent Edges (HotEdgeVideo '19), pp. 21-26 (2019)
      [PDF (Open Access)] [Code]
  • Journal Articles
    • Y. Matsubara, D. Callegaro, S. Baidya, M. Levorato and S. Singh, "Head Network Distillation: Splitting Distilled Deep Neural Networks for Resource-constrained Edge Computing Systems", IEEE Access, (2020)
      [Early Access] [Code]