Ryoya Yamasaki Ryoya Yamasaki

Assistant Professor, Hitotsubashi Institute for Advanced Study, Hitotsubashi University *Profile is at the time of the award.

2026HagukumuScience & Engineering(hagukumu)

researchmap

Research topics: Unimodality-Promoting Regularized Learning for Ordinal Data

Keyword

Summary: Ordinal regression addresses conditional probability prediction and classification for oridnal data where the underlying target variable possesses a "natural ordinal relation" (e.g., Agree / Neutral / Disagree)._x000D_
In my previous works, I discovered that in many ordinal data, the conditional probability distribution would be unimodal across a broad domain of the explanatory variable; even where it would not be strictly unimodal, it would remain nearly so._x000D_
In this work, aiming to improve prediction performance for small-size ordinal data, I will construct a systematic regularized learning framework that induces unimodal or near-unimodal conditional probability distributions.

Message

By focusing on the inherent "unimodality" of ordinal data, I aim to establish a novel learning methodology that enables high-precision predictions even with small-size datasets. Through this grant, I hope to contribute to improving the accuracy of diverse real-world applications, such as survey analysis and rating predictions.