Theory of Programming and Artificial Intelligence
We conduct a wide range of research on basic theory and applications of computational intelligence, mathematical informatics, and software engineering. Specifically, we promote research on machine learning algorithms, mathematical programming, distributed algorithms, software measurement and analytics, mining software repositories, human behavior analysis, human-machine interaction, computer vision and so on.
Development of distributed algorithms for multiagent systems
I am studying methods for multiple agents to solve problems in a distributed and cooperative manner. Specifically, I design distributed algorithms for agents to solve given large-scale problems efficiently under the situation that each agent can communicate only with a small number of neighboring agents, and apply them to principal component analysis, nonnegative matrix factorization, and federated learning.
Development of software project management simulator
In this study, we are developing a simulator for learning quantitative software development management (https://www.okayama-u.ac.jp/user/salab/spm/). The simulator enables us to learn effort estimation, developer assignment considering their skills, schedule and cost management considering productivity factors, and engineer skill development through on-the-job training.
Human behavior analysis
Today, artificial intelligence systems are becoming more and more pervasive in our daily lives. In that respect, the ultimate goal of this laboratory is to promote human-machine interaction and develop human-friendly technology by understanding human behavior. To this end, we are conducting research in various research areas of artificial intelligence, such as head pose estimation, eye detection and tracking, joint attention, and robot learning.
Image understanding via inverse rendering
Computer graphics algorithms generate 2D images from 3D parameters, such as shape and pose of each object in a 3D scene, considering laws of optics and geometry. Its inverse problems analyze 2D images to reconstruct 3D parameters, which enables computers to understand or accurately measure objects in a 2D image. I am studying efficient numerical optimization methods required for such inverse problems.