Masatoshi Nagano
Papers
5
Total Citations
48
H-Index
4
About
Masatoshi Nagano is a researcher whose work sits at the intersection of robotics, machine learning, and cognitive science, focusing on how machines can autonomously learn from continuous, high-dimensional data streams. His core research area is unsupervised segmentation—the challenge of enabling robots to break down raw sensory input, such as motion or speech, into meaningful, reusable units, much like humans do when learning words or actions. Nagano’s major contributions include developing a suite of hierarchical probabilistic models, most notably HVGH (2019, 20 citations), which combines deep neural compression with statistical generative models to segment complex time series. This work builds on his earlier GP-HSMM framework (2018, 15 citations) and extends into motion segmentation (2019, 6 citations) and spatio-temporal categorization for first-person-view robot videos (2022, 4 citations). His most recent work (2023, 3 citations) tackles the ambitious goal of unsupervised phoneme and word acquisition from continuous speech, directly modeling the double articulation structure of language. By advancing methods that allow robots to discover structure without labels, Nagano is laying the groundwork for more autonomous, human-like learning in artificial systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5