For the sake of viewer convenience, the content is shown below in the alternative language. the link to switch the active language.
Hello, this is Shimizu from Public Relations.
In the blink of an eye, there are only a few days left in this year.
The Matsuo Lab has been working hard on research, education, social implementation, and awareness of deep learning this year, and the number of engineers and administrative staff has increased, making the lab a very lively place!
Therefore, I would like to take a few moments to reflect on this year’s Matsuo Institute.
There were many releases in the first half of the year!
So here are again four topics that have been discussed both inside and outside the Matsuo Institute this year.
1. Publication of Japanese translation of “Deep Learning” by Goodfellow et al.
The Matsuo Lab has taken the initiative in translating a textbook written by Goodfellow and Bengio, who are well-known researchers in deep learning, which can be considered the definitive textbook for studying deep learning.
You were also a topic of discussion at the preliminary online publication (now closed to the public) in 2017.
Dr. Iwasawa, a specially-appointed assistant professor who coordinated the Matsuo Lab’s translation team, also commented on the project.
Project Assistant Professor Yusuke Iwasawa
This book was published with the support of a translation team consisting mainly of members who had participated in study groups organized by the Matsuo Lab and many others. Although DL research is advancing rapidly, its foundation is supported by many years of research on linear algebra, machine learning, and basic neural nets, which are the subject of this book. We hope this book will help you to acquire such basic knowledge.”
Click here to purchase the book .
The reading material is available at “Deep Learning.jp”, a study group run by Matsuo Lab.
2. Educational activities such as “DL4US” and “Free Release of Classroom Exercise Content
Two free exercise contents of Matsuo Lab’s popular courses are now available. What? You can study this for free? and “You can study this for free?
◆GCI Data Scientist Training Course
The contents of the data analysis part of the Global Consumer Intelligence Endowed Course at the University of Tokyo.
Deep Learning Basic Course
Exercise contents used in the Deep Learning Fundamentals Course being held at the University of Tokyo. Although linear algebra and machine learning are required as prerequisite knowledge, it is designed so that students can learn important knowledge for constructing new Deep Learning models and conducting advanced research and development from the basics.
We also opened the first and second terms of DL4US, a free, application-oriented online education program aimed at training advanced Deep Learning engineers, this year.
The program has received media coverage and attracted more than 1,900 applicants. It was an event that made us realize that interest in Deep Learning and desire to improve technical skills is growing in society beyond what we had imagined.
Dr. Nakayama, who leads this team and has spoken at various seminars on AI human resource education, commented on the event.
Kotaro Nakayama, Research Director
The Matsuo Lab conducts large-scale classes for students and working professionals with the aim of developing data scientists and personnel with mastery of Deep Learning.
The DeepLeanring Basic Course and the Data Science Course, which were released earlier this year, have received a great response, especially on social networking services. We have also expanded the scope of our activities while boldly taking on new challenges, such as expanding the scale of our data scientist training course for working professionals to 300 students.
However, we feel that we have not yet been able to provide sufficient education, both in comparison with overseas universities and in consideration of the current importance of AI human resources. We intend to put more effort into our activities in the future as one of the activities of our laboratory.”
3. Papers
The Matsuo Lab started with web engineering research with the goal of “creating ‘intelligence’”, and in recent years, deep generative modeling and deep reinforcement learning have become the pillars of our research area.
This year, our paper on domain manipulation using deep generative models applied to speech synthesis was accepted for publication in Interspeech2018. Mr. Akuzawa, who wrote that paper, commented on it.
Kei Akuzawa, 2nd year master’s student
Deep learning is a rapidly evolving research area, and the Matsuo Lab is working hard to keep up with the latest developments by hosting readings of papers and implementation meetings.
We have many participants from various laboratories and companies, not only from our laboratory, and we find it very useful to get a bird’s eye view of various research areas, such as image, speech, reinforcement learning, and theory.
In addition to the publication of the paper, Researcher Suzuki released “pixyz,” a library for developing deep generative models, and B4 Taniguchi released a reproduced implementation of DeepMind’s proposed GQN model, which attracted attention. I felt that our research activities were bearing fruit.
All the faculty members, researchers, and students involved in basic research are highly motivated to produce more impactful research results next year.
4. “Deep 30” venture capital to update the future together with the limitless possibilities of deep learning
Matsuo Lab has produced entrepreneurs such as PKSHA Technology and Gunousy, and this year we established a venture capital firm, Deep30 Investment Limited Partnership (“Deep30”), with the aim of investing in seed-stage startups that could revolutionize the world through the use of deep learning technology. Deep30 Investment Limited Liability Partnership (“Deep30”) was established to invest in seed-stage startups that can revolutionize the world by utilizing deep learning technology.
Mr. Tazoe, a representative member of the association, who has a career in business, having worked for a major consulting firm and listed on a venture company’s stock exchange, commented, “I am very pleased to be here.
Academic Support Specialist / Deep30 Investment Limited Liability Partnership Satoshi Tazoe, Representative Partner
Deep30 is a venture capital firm that invests in companies with strengths in deep learning.
This initiative itself can be described as a venture that was started through discussions with Dr. Matsuo and many other researchers and staff members on issues that I had been thinking about in my daily work at the university. I feel a good flow of support for those who believe in the broad possibilities of technology and take the initiative to take action. Of course, there are many challenges and things that are not yet in place.
We will continue to strengthen industry-academia collaboration and support for ventures that will take off in the coming year and beyond, so that we can become the blood of the ecosystem for a series of basic research and venture creation.
In the coming year, all members of the laboratory will continue to work on research, education, social implementation, and education of artificial intelligence to society in order to “create intelligence and open up the future”.
All of us at Matsuo Lab wish you good health and continued growth for your company.
Please have a happy New Year.
We look forward to working with you in the New Year.