ABOUT
Ryo Yuki (CV), (Twitter), (LinkedIn)
Doctoral Course of Information Science and Technology
Department of Mathematical Informatics
The University of Tokyo
Yamanishi Lab.
Email: jie-cheng-ling[at]g.ecc.u-tokyo.ac.jp
I am a Machine Learning Expert with three years of combined research and work experience. I have a strong sense of responsibility and good communication skills, both of which have been cultivated by my experiences as a team leader. I also have much programming experience through developing time-series analysis softwares and study of machine learning.
EDUCATION
Mar. 2024 (Expected) Ph.D. in Information Science and Technology, The University of Tokyo, Tokyo.
Faculty Mentor: Professor Kenji Yamanishi.
Mar. 2021 M.E. in Information Science and Technology, The University of Tokyo, Tokyo.
Faculty Mentor: Professor Kenji Yamanishi.
Dissertation: Change Sign Detection with Two-Stage MDL Change Statistics.
Mar. 2021 B.E. in Precision Engineering, The University of Tokyo, Tokyo.
Faculty Mentor: Associate Professor Yutaka Ohtake.
Dissertation: A Machine Learning Approach for Fast X-ray Computed Tomography Scan by Deblurring Transmission Images (in Japanese).
RESEARCH INTEREST
・Graph Mining
・Statistical Model Selection
・Anomaly Detection
・X-ray CT
KEY SKILLS
Machine Learning Knowledge: Advanced
Mathematics: The Statistical Certification, Grade 1 (in Japan) and Mathematics Certification Level 1st (in Japan).
Python,C,C++: Advanced
Java,Ruby,HTML and CSS: Beginner
Git: Advanced
Unix Commands: Advanced
Arduino,CAD: Intermediate
Languages: HSK V 204 pts and HSKK Beginner (Chinese)
RESEARCH EXPERIENCE
Apr. 2021 - Sep. 2022 Dimensionality Selection of Hyperbolic Graph Embeddings using Decomposed Normalized Maximum Likelihood Code-Length
Faculty Mentor: Professor Kenji Yamanishi
• Research Topic: dimensionality selection of hyperbolic graph embeddings using decomposed normalized maximum likelihood code-length, which is a variant of minimum description length principle.
• Accepted for publication in 22nd international conference on data mining (ICDM2022) (slides).
Apr. 2019 - Mar. 2021 Change Sign Detection with Two-Stage MDL Change Statistics
Faculty Mentor: Professor Kenji Yamanishi
・Master’s study, dissertation.
・Research Topic: detection of change signs in data streams by two-stage change detection approach.
Feb. 2020 - Jul. 2020 Change Sign Detection with Differential MDL Change Statistics and its Applications to COVID-19 Pandemic Analysis (link)
・Research Topic: detection of change signs in data streams using differential information of change scores and applications to COVID-19 pandemic analysis.
・Implemented an online alerting system of COVID-19 pandemic analysis (link).
Apr. 2018 - Mar. 2019 A Machine Learning Approach for Fast X-ray Computed Tomography Scan by Deblurring Transmission Images (link)
Faculty Mentor: Professor Yutaka Ohtake
・Bachelor’s study, dissertation.
・Research Topic: acceleration of precise CT scan by deblurring transmission images using convolutional neural networks.
・Accepted by 10th conference on industrial computed tomography 2020 (iCT2020) (slides).
INTERNSHIP EXPERIENCE
Aug. 2021 - Sep. 2021 Preferred Networks, Inc, Japan.
• Worked on change detection and its variable selection in multivariate time-series data (link).
• Investigated and implemented additive Hilbert-Schmidt independence criterion (aHSIC, Yamada et al., 2013), which is based on the kernel method and Lasso (Tibshirani, 1996 and 2011).
• Implemented the following improvements to the algorithm: (1) searching for multiple points in a window, (2) extension of aHSIC to handle changes that gradually occur, and (3) automatic selection of the regularization factor λ.
• Improved the performance of the method in terms of area under the curve (AUC) and mean average precision (MAP).
Sep. 2018 - current pluszero, Inc, Japan.
・Developed an automatic electricity consumption prediction system for a company with AWS sagemaker. Managed the development flow as a team leader and implemented the machine learning system with python as a developer. Improved the accuracy and time efficiency of prediction.
・Developed a time-series analysis software for one of the biological laboratories at the university of Tokyo. Managed the development flow as a team leader and implemented the system by python. The processing time for each data sequence was shortened from 100 minutes to 5 minutes with the developed software, which improved the efficiency of biological research.
・Developing an autonomous time-series analysis software for a company, which is motivated by ”AutoML”, a machine learning library that chooses suitable machine learning methods and adjusts the hyperparameters automatically. Investigated time-series algorithms proposed in previous studies, assigned their implementation to team members, and implemented 5 algorithms from scratch.
Aug. 2019- Sep. 2019 Wantedly Inc.
・Improved the recommendation system of Wantedly Visit, a website that recommends job applicants to companies and vice versa.
・Implemented an autonomous detection system of anomalous users and companies with python, which try to boost their recommendation ranking of recruitments intentionally by automatic accessing.
Aug. 2016 - Sep. 2017 Morpho, Inc, Japan.
Developed an Android camera software and researched about super-resolution of human facial pictures.
Nov. 2015 - Mar. 2018 Hanamaru lab Inc, Japan.
Developed programming related to teaching materials for schoolchildren. The number of customers is 5 times more than the previous year by my development and developed an intellectual training software as a member of a team.
AWARDS
Feb. 2022 Excellent Student Award in the ”Chinese Bridge” Online Program, Zhejiang University
Aug. 2019 AI RUSH 2nd place
A machine learning competition conducted by LINE Corporation. 2nd place out of 100 teams.
Jul. 2018 Dept. of Precision Eng. "English Presentation Program" Best Overall
CERTIFICATIONS
Dec. 2019 HSK V 204pts
A certification that measures Chinese language ability authorized by the Chinese government. HSK includes listening, reading, and writing tests.
Oct. 2019 Mathematics Certification Level 1st
A certification that measures mathematical skills.
Mar. 2019 HSKK Beginner
A certification that measures Chinese language ability authorized by the Chinese government. HSKK includes speaking tests.
Dec. 2018 The Statistical Certification, Grade 1
A certification that measures mathematical ability to analyze data.
PROJECTS
Sep. 2016 - May.2018 Manager of Seimitsu Lab. from the Dept. of precision engineering
Seimitsu Lab. is a project which focuses on programming and mathematics related to engineering. We showed VR games, autonomous cocktail makers, projection mappings, and more projects. I managed each project as the responsible person and created the application which suggests the most preferable hairstyle using machine learning as a developer.
We took interviews with several media. The articles are (1), (2), (3), and here (in Japanese).
Jan. 2017 - Dec. 2017 Choir Shirobara Manager
Managed a chorus club for a year. Mainly, increased the number of members for better club activity.
SOFTWARES
May. 2018 Hairstyle suggester by CNNs. Github (in Japanese)
This application suggests the optimal hairstyle by inputting the human face image. Its algorithm mainly utilizes CNN(Convolutional Neural Network) which is one method of deep learning. Implemented by Keras.
Jun. 2017 TSP Visualizer by GA.
Github (in Japanese)
Made for the class which I enrolled in. TSP is a problem that is hard to solve correctly. Finding an approximate solution by GA (Genetic Algorithm). Visualized by OpenGL.
PUBLICATIONS
Yuki, R., Ike, Y., & Yamanishi, K. (2022). Dimensionality selection of hyperbolic graph embeddings using decomposed normalized maximum likelihood code-length. accepted for presentation at ICDM2022 (regular paper).
Yamanishi, K., Xu, L., Yuki, R., Fukushima, S., & Lin, C. H. (2021). Change sign detection with differential MDL change statistics and its applications to COVID-19 pandemic analysis. Scientific reports, 11(1), 1-15. (link)
Yuki, R., Ohtake, Y., & Suzuki, H. (2022). Acceleration of X-ray computed tomography scanning with high-quality reconstructed volume by deblurring transmission images using convolutional neural networks. Precision Engineering, 73, 153-165. (link)
Yuki, R., Ohtake, Y., & Suzuki, H. (2020). Deblurring X-ray transmission images using convolutional neural networks to achieve fast CT scanning. The 10th international conference on industrial computed tomography (iCT2020), Wels, Austria, Feb. 2020 (full paper). (link)