-
🌱 My research interests are ML, DL, and NLP, especially multimodal in various domains such as language, images, audio, and tabular.
-
📫 How to reach me [email protected], Google Scholar
NEC Corporation, Tokyo Japan (April 2024- Present)
- Full-time
- AI/ML Researcher @ Biometrics Research Laboratories
- NEC github account: https://github.com/somaoni
IDAJ Co. LTD., Kanagawa Japan (April 2023- March 2024)
- Part-time
- AI/ML Engineer
- Creation of physical simulation models
- Acceleration through ML replacement of physical simulation models
Preferred Networks Inc., Tokyo Japan (October 2022- March 2023)
- Part-time
- AI/ML Researcher
- Presented at the poster session of IBIS2022.
- Proposal of pre-trained models for tabular data
- TabRet: Pre-training Transformer-based Tabular Models for Unseen Columns (link)
- Accepted at the Workshop on Understanding Foundation Models at ICLR 2023.
- Code: https://github.com/pfnet-research/tabret
- The proposed method can be applied to data with different distributions during pre-training and fine-tuning in tabular data.
- TabRet: Pre-training Transformer-based Tabular Models for Unseen Columns (link)
Preferred Networks Inc., Tokyo Japan (August 2022- September 2022)
- Intern
- AI/ML Researcher
- Presented the research outcomes from the summer intern (2 months) at an internal company presentation.
GHELIA Inc., Tokyo Japan (October 2021- July 2022)
- Intern
- AI/ML Engineer
- Developed a plug-in for DL application OSS (1 month).
- Research and development on efficient anomaly detection of graphic bugs in games (3 months).
- Research and development on frame interpolation for animation (3 months).
- Research and development on anomaly detection of rust and missing bottles in bridges (1 month).
-
Why Do Tree Ensemble Approximators Not Outperform the Recursive-Rule eXtraction Algorithm? (link) [March 2024]
-
Rethinking Data Augmentation for Tabular Data in Deep Learning (link) [May 2023]
- arXiv.
- Code: https://github.com/somaonishi/MTR
-
TabRet: Pre-training Transformer-based Tabular Models for Unseen Columns (link) [March 2023]
- Accepted at the Workshop on Understanding Foundation Models at ICLR 2023.
- Work done during internship at Preferred Networks, Inc.
- Code: https://github.com/pfnet-research/tabret