Kento Nozawa


01 April 2025

Short bio

Kento Nozawa is an engineer at Preferred Networks, Inc. Recently, he has fine-tuned in-house LLMs, PLaMo. He completed his Ph.D. under the supervision of Dr. Issei Sato at Issei Sato Lab in The University of Tokyo. Back then, he was working on self-supervised representation learning, especially contrastive representation learning.

Selected papers

  1. Han Bao, Yoshihiro Nagano and Kento Nozawa. On the Surrogate Gap between Contrastive and Supervised Losses. In ICML, pages 1585–1606, 2022. paper, video, code, poster, arXiv.Alphabetical ordering and equal contribution.
  2. Kento Nozawa and Issei Sato. Evaluation Methods for Representation Learning: A Survey. In IJCAI-ECAI Survey Track, pages 5556–5563, 2022. paper, video, slides.Extended version: ``Empirical Evaluation and Theoretical Analysis for Representation Learning: A Survey’’. 2022. arXiv
  3. Kento Nozawa and Issei Sato. Understanding Negative Samples in Instance Discriminative Self-supervised Representation Learning. In NeurIPS, pages 5784–5797, 2021. paper, slides, code, poster, arXiv.
  4. Kento Nozawa, Pascal Germain and Benjamin Guedj. PAC-Bayesian Contrastive Unsupervised Representation Learning. In UAI, pages 21–30, 2020. paper, video, slides, code, arXiv.
  5. Atsunori Kanemura, Yuhsen Cheng, Takumi Kaneko, Kento Nozawa and Shuichi Fukunaga. Imputing Missing Values in EEG with Multivariate Autoregressive Models. In EMBC, pages 2639–2642, 2018.

Pre-print

  1. Kento Nozawa, Takashi Masuko and Toru Taniguchi. Enhancing Large Language Model-based Speech Recognition by Contextualization for Rare and Ambiguous Words. 2024. arXiv.

The full publication list is available on Google Scholar.