Efficient Inference for Flow Matching via Unified Path CFG
Kosuke Ukita, Tsuyoshi Okita
The 30th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD 2026)
Kosuke Ukita
浮田 嵩祐
MSc student / Researcher
Kyushu Institute of Technology
Fukuoka, Japan
Department of Creative Informatics, Graduate School of Computer Science and Systems Engineering.
こんにちは．浮田 嵩祐です．AIに関する研究をしている大学院生です． 生成AIの効率化や基盤モデルの構築，表現学習などの深層学習領域や，機械学習の基礎理論(e.g. Grokking)に関心を持っています． 最近はFlow Matchingをよく触っています．
News
2026-02-08我々の論文が国際会議PAKDD2026にOral presentationとして採択されました．
2026-01-23我々の論文が国際会議AISTATS2026にSpotlight presentation (Top 3%)として採択されました．
Publications
Recent 3 publications
High-Performance Self-Supervised Learning by Joint Training of Flow Matching
Kosuke Ukita, Tsuyoshi Okita
The 29th International Conference on Artificial Intelligence and Statistics (AISTATS 2026)
Foundation Models to Tackle Activity Recognition in Unknown Domain: Sussex-Huawei Locomotion Challenge 2025 Task 2
Tsuyoshi Okita, Kosuke Ukita, Asahi Miyazaki, Daichi Kubota, Jukichi Ota, Naoki Kagiyama, Asahi Nishikawa, Nagayasu Daichi, Syunya Tomitaka, Daisuke Nozaki, Yuki Odo, Raku Yamashita, Xiaolong Ye, Huayu Gao, Kazuki Okahashi, Koki Matsuishi, Masaharu Kagiyama, Kodai Hirata, Haruki Kai, Lin Wang, Hristijan Gjoreski, Mathias Ciliberto, Paula Lago, Kazuya Murao, Daniel Roggen
Companion of the 2025 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp Companion '25), pp. 983–991
October 2025 Espoo, Finland
OralRefeered
Awards
Spotlight Presentation (Top 3%)
The 29th Annual Conference on Artificial Intelligence and Statistics (AISTATS 2026)May 2026
Incentive Award (奨励賞)
IPSJ Kyushu Branch (情報処理学会九州支部)March 2024
Highest Award (最優秀賞)
The 8th Sports Data Analysis CompetitionMay 2019
Others
Research Interests
Generative AIFlow MatchingFoundation ModelsMultimodal LearningDiffusion ModelsMachine Learning
Languages
PythonC++CJavaJavaScriptTypeScriptGoSQLR
Frameworks & Tools
PyTorchTensorFlowHugging FaceJupyterABCIDockerGit
Other Skills
Deep LearningNeural NetworksModel TrainingData ProcessingAPI DevelopmentLinux