March 4, 2024

Becoming a repeat winner at top international conferences through collaboration between industry, government, and academia

Senior Researcher, Data Science Laboratories
Ryota Higa

After leaving a doctoral program in condensed matter theory research without completing his degree, Higa joined NEC in 2016. He has been involved in researching deep reinforcement learning and automated negotiation AI utilizing such techniques, and his research papers have been repeatedly accepted by AAAI-22, 23, and 24, a top international conference in artificial intelligence. In addition to also working at the National Institute of Advanced Industrial Science and Technology (AIST), RIKEN, and BIRD INITIATIVE, Inc., Higa has continued his vigorous research activities working as an expert investigator at the National Institute of Science and Technology Policy within the Ministry of Education, Culture, Sports, Science and Technology (NISTEP). He is also drafting new research themes and launching new teams.

The ability to conduct graduate school-level research while working at a company

After leaving a doctoral program, I joined NEC and still need to complete my degree. At the time, I was not getting research results and was unsure whether I should remain in academia a little longer or enter the workforce and look for a different path. As a result, I tried out the role of a corporate researcher, which was new to me, and when I considered once again which path was right for me, I decided to join NEC.
I was interested in NEC because it also has a presence in academia, as many of its research papers have been published. It has many famous researchers both in Japan and overseas. I got the impression that even among Japanese companies, NEC is very dedicated to research, so I applied. Later, when I could tour the laboratories during the screening process, I was surprised to learn they are engaged in graduate school-level research. Speaking of corporate laboratories, I thought they were mainly involved in "development" by tweaking existing technologies to improve them. Still, when I learned that NEC is conducting full-scale research, I became even more eager to join the company. Another vital motivating factor for joining the company was the global atmosphere, with researchers from Japan and various other countries.
Fortunately, I was able to join the company, but my original expertise was in condensed matter theory. To get acquainted with machine learning, I entered a study period of about one to two years. Specifically, after joining the company, I was assigned to Fujimaki's team, which was later carved out and established as dotData, Inc. , so I was constantly amazed by the level of excellence of my colleagues. However, I was told then that "many people from theoretical physics and physics are actively working in this area." At the time, I wondered if that was true, but I now know that is true. The mathematical models and other skill sets used in physics can often be used in a similar context in machine learning and artificial intelligence areas. Therefore, as I started to be able to translate between physics and machine learning in my head effectively, the transition instantly became smoother. People with a physics background are instead citing the techniques in the areas that they learned to put them to use in artificial intelligence and machine learning. I hope that students majoring in physics will take on the challenge of research in artificial intelligence and machine learning.

Research papers accepted by AAAI, a top international conference in artificial intelligence

My concurrent posting at NEC-AIST AI Cooperative Research Laboratory starting in 2019 was a significant turning point in advancing my research. I was invited to participate in research on automated negotiation AI (artificial intelligence that mutually adjusts interests), which was launched as a national project (Cross-ministerial Strategic Innovation Promotion Program) and is now starting to produce visible results. From 2020, I continued to post research papers at a pace of about one paper every three months. After my research paper was accepted in 2022 by AAAI, a top international conference in artificial intelligence, my research papers have continued to be received every year since. My work has also been recognized on an academic level.
The kind of research that we are pursuing cannot be realized by NEC alone. We do not simply create theories, as we can conduct our research only when we visit customers in the logistics and manufacturing industries to receive data and collaborate with various external stakeholders. In addition to collaborating with AIST, I have held joint positions with organizations such as RIKEN and BIRD INITIATIVE, Inc., a company that promotes carve-outs with financing from various companies, including NEC and other shareholders. These are essential initiatives for creating new technologies and forms of value by combining multiple organizations' business and technology domains around my core specialty of deep reinforcement learning.
I continue to engage in weekly discussions with university professors, researchers, and students, as well as international conferences. This environment suits me to an extremely high degree. I feel that I can conduct research more rigorously than when I was in university, and I can also receive live data from customers and ask for help with any problems, so I feel a sense of purpose in pursuing research not for its own sake but research that leads to businesses that contribute to society.
In addition, when you present many times at international conferences, well-known professors overseas will invite you to engage in joint research, etc., which creates various connections. The other day, someone at NISTEP asked me to help create a report, and I contributed individually. The ability to be chosen by name as a corporate researcher is unique to NEC.

At NEC, you can immediately work at the forefront after joining the company.

As I mentioned at the outset, I joined the company after thinking I would experience corporate research and decide whether to proceed with academia or corporate research. Now, I am glad I chose to join NEC and hope to continue researching this way. To begin with, if you look around the world, "GAFAM" and other famous companies are leading research in the artificial intelligence field. Moreover, companies are more generous than academia regarding human and economic support for research. In artificial intelligence and machine learning research, one good option is to try joining a company.
Among many companies, NEC is focusing its efforts on research and has developed an environment that is highly conducive to young people. The laboratories contain diverse human resources and know-how built over many years, and there is also a broad network with ties to client companies, academia, etc. We certainly aspire to the flashy results of companies such as GAFAM. Still, the appeal of NEC is that it has an environment where one can independently conduct research on the front lines several years after joining the company. Many young researchers produce results at top conferences at NEC. Everyone is trying hard to catch up to and surpass them. The combinational optimization and deep reinforcement learning I have patiently been working on since joining the company has been recognized as a new theme. I will be appointed as a technology leader in the DS Lab in FY 2024. In addition, I will also be involved in launching a new team project as part of the NEC-AIST collaboration. I will also continue my activities to produce academic results and work for the company's benefit.
Moreover, even if you try corporate research and want to follow a different path outside of research, that is okay. At the NEC laboratories, there are various career paths where you can be accepted if, for example, you realize that you are better suited to engineering or feel that engaging in business development with customers is enjoyable. Many people are actively working as valuable engineers who can understand the core technologies to a deep level, and some people design solutions by combining the optimal technologies based on customer issues. Flexible career paths like these are unique to large companies. When looking at various news outlets, start-up companies seem spectacular. Still, another path is to effectively use large companies' deep pockets as a place for career development.

A day at work

Message to my past self in my school days