Research Engineer
Contact: shuhei.watanabe.utokyo [at] gmail.com
I am a research engineer at the AutoML group of Preferred Networks Inc. Before joining the group, I was studying under the supervision of Prof. Frank Hutter at the University of Freiburg in Computer Science department. The main interests of my research are practical extensions of blackbox optimization on expensive functions. More specifically, I have been tackling multi-objective optimization, constraint optimization, meta-learning of hyperparameter optimization, and the interpretation of optimizations. For more details, please check my Research Experiences.
Prior to the Master course, I finished the Bachelor at the University of Tokyo, Systems Innovation Faculty supervised by Prof. Chen Yu in March 2020. During my Bachelor, I was engaged in the National Institute of Advanced Industrial Science and Technology (AIST) supervised by Masaki Onishi. The details of my works are available here (My works). Personal ongoing projects are German and functional analysis learning.
I was intensively working on extensions of tree-structured Parzen estimator (TPE). For more details, please check my publications.
I listed the ones related to only research. For more details, please check my CV.
AutoML 2023 Travel Awards (500 EURO).
NeurIPS 2022 Complimentary Registration (350 USD) awarded by Gaussian Processes workshop organizers.
ELIZA MSc scholarship (1054 EUR/Month) selected by ELLIS/ELIZA unit Freiburg.
4 students were selected from the whole Computer Science Master Program in the University of Freiburg.
Deutschlandstipendium (300 EUR/Month for 12 months).
ITO Foundation of International Education Exchange (2000 USD/Month for 2 years).
The acceptance rate was about 6.2% (= 12/193).
Hatakeyama Award from the Japan Society of Mechanical Engineers for the distinctive grades at the mechanical engineering related faculties at the University of Tokyo.
The acceptance rate was about 1.5% (= 5/340).
PRMU 2018 Yearly Research Encouragement Award for the Paper Speed Up of Hyperparameter Tuning with Nelder-Mead Method by Parallel Computing.
The acceptance rate was about 1.8% (= 3/170).