Research

Tomoharu　NAGAO

Tatsunori MORI

We are conducting research on a system that enables users to easily utilize information expressed in words, targeting a large volume of documents. Our goal is to enhance intelligent information access systems based on natural language processing technologies, such as question answering, automatic summarization, information extraction, and information retrieval. In recent years, we have been particularly focusing on the utilization of large language models. For example, in a methodology we named "Think from Words," we demonstrate that by making the language model consider the meanings of individual words, the processing accuracy for the entire sentences that include those words improves. Additionally, through collaborative research with companies, we are exploring a system that uses large language models to extract, summarize, and visualize knowledge from documents that contain the expertise accumulated by those companies.

Katsunori OKAJIMA

Masanori YASUMOTO

Recent research：

  • Management of open shared platform technologies
  • Firms' innovation strategies related to open technologies
  • Governance of community and ecosystem for open technologies and data
  • Structural analysis of technology and knowledge inside and outside of firms, etc

Naoshi TOMII

Toshihiko SHIRAISHI

Keisuke SHIMA

Humans usually reach out and grab distant objects, and casually perform various movements such as walking and running without falling. However, in order to perform these movements naturally, it is necessary to contract a large number of muscles in a well-balanced manner and perform them with appropriate force and timing. In addition, humans can acquire various information by actions such as seeing something with their eyes or feeling it with their skin, and can make various judgments and predictions based on the current situation and past experience．
In our laboratory, we elucidate the principle of the mechanism that realizes the skillful movement of humans and apply intelligent robot technology equipped with artificial intelligence that thinks and judges flexibly like humans, and effectively humans in various scenes. We are conducting research activities with the aim of supporting humans．

  • Neural network model that realizes probabilistic pattern recognition
  • Unlearned class estimation model with complementary event distribution
  • Pulse neuron model considering hardware implementation
  • Evolved neuron chip using FPGA

Shinichi SHIRAKAWA

Research topic：

  • Evolutionary Computation and Black-Box Optimization
  • Machine Learning and Deep Learning
  • Applications (e.g., Computer Vision, Medical and Healthcare Data, IoT, etc.)

Please see the publication page about our recent research topics.

Ryoji TANABE

I am interested in improving the fundamental understanding and applicability of evolutionary algorithms through an experimental approach. For example, I prefer to analyze the behavior and performance of evolutionary algorithms through benchmarking. I proposed some new components of evolutionary algorithms based on the analysis results.

Tomoki HAMAGAMI

Manabu KUROKI

Chika SUGIMOTO

Hiroaki GOTO

Masaya NAKATA

Expensive optimization

Many real-world applications require optimizing expensive-to-evaluate objectives, as objective values are evaluated often with computationally expensive simulations or costly experiments. An example is the optimization of aircraft wing using CFD simulation, which is computationally expensive. To solve such expensive optimization problems, we explore sample-efficient optimization techniques. Especially, high-dimensional multi-objective problems are our main focus to be studied.

Evolutionary machine learning

Integrating meta-heuristics into machine learning approaches bring highly adaptive and autonomous ability, such as auto-design of learning models, auto-tuning of hyper-parameters, and extraction of explainable models. We explore evolutionary machine leaning techniques; evolutionary rule-based learning, evolutionary symbolic regression, evolutionary neural architecture search, and etc. We are interested in realizing the interplay between evolution and learning on a computer intelligence scheme.