Ryota Tomioka

Ryota Tomioka is a machine learning researcher at Microsoft Research whose recent work applies deep learning to materials and proteins. After a PhD at the University of Tokyo in 2008, he worked there as an assistant professor, spent time at the Toyota Technological Institute at Chicago as a research assistant professor, and joined Microsoft Research Cambridge in 2015. In 2016 he coauthored f-GAN, which trains generative neural samplers with f-divergences, and QSGD, which compresses gradient updates for parallel training. MatterGen, a diffusion model for inorganic materials described in a January 2025 Nature paper, came from research he led with Tian Xie, and he is among the authors of the BioEmu-1 protein model presented in February 2025. He is a Senior Principal Research Manager at Microsoft Research AI for Science.

Now at: Microsoft Research
Works: 116
Citations: 8,884
h-index: 35
Badges: Bridge

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

2008

PhD at the University of Tokyo

He receives his PhD from the University of Tokyo in 2008, with Kazuyuki Aihara as thesis advisor. During his doctoral studies, he is a visiting research associate at the Technical University of Berlin.
2008–2011

Assistant professor in Tokyo

In April 2008, he moves to Masashi Sugiyama's group at the Tokyo Institute of Technology as a postdoctoral researcher, and in 2009 he joins Kenji Yamanishi's group in the Department of Mathematical Informatics at the University of Tokyo as an assistant professor. In 2011, he coauthors the chapter "Augmented Lagrangian Methods for Learning, Selecting, and Combining Features" in the MIT Press volume Optimization for Machine Learning.
2013–2015

Research assistant professor at TTIC

He joins the Toyota Technological Institute at Chicago as a research assistant professor in October 2013. In 2015, he coauthors "Norm-Based Capacity Control in Neural Networks" with Behnam Neyshabur and Nathan Srebro, a study of the capacity of norm-constrained feed-forward networks.
November 2015

Joining Microsoft Research Cambridge

He joins Microsoft Research Cambridge as a Researcher in November 2015. His research lies in the intersection of machine learning, optimization, and neural networks.
2016

f-GAN and QSGD

In June 2016, he coauthors f-GAN with Sebastian Nowozin and Botond Cseke; the paper shows that any f-divergence can be used for training generative neural samplers. In October 2016, he coauthors QSGD with Dan Alistarh, Demjan Grubic, Jerry Li and Milan Vojnovic, a family of compression schemes for gradient updates that trains a ResNet-152 on ImageNet 1.8 times faster to full accuracy on 16 GPUs.
2023–2025

MatterGen for materials design

In December 2023, Microsoft Research AI4Science announces MatterGen, a diffusion model for generating novel, stable materials, and he is one of the authors of the announcement. In January 2025, the Nature paper "A generative model for inorganic materials design" appears; he and Tian Xie lead the research and are its corresponding authors.
February 2025

BioEmu-1 for protein structures

In February 2025, Microsoft Research presents BioEmu-1, a deep learning model that generates thousands of protein structures per hour on a single graphics processing unit, and he is among its authors.
As of October 2026

Research manager at AI for Science

He is a Senior Principal Research Manager at Microsoft Research AI for Science. His research interest is learning algorithms for atomic simulations.