Ikko Yamane: “The interdisciplinarity at ENSAI is the best environment for building collaborations” 

Ikko Yamane joined ENSAI as an Assistant Professor in Computer Science last fall. The University of Tokyo PhD graduate teaches various courses while carrying his research in Machine Learning. He tells us about his experience over the last few months, working in a French Grande Ecole.

Yamane is a member of the CREST research laboratory, a mixed research unit of the CNRS attached to the Group of National Schools in Economics and Statistics  (GENES) and Ecole polytechnique.

From Tokyo to Rennes

I did my Bachelor’s and Master’s degrees at Tokyo Institute of Technology (soon to become part of the Institute of Science Tokyo once it merges with Tokyo Medical and Dental University). I studied at the University of Tokyo for my PhD.

My research career started with the design of a decoding algorithm but switched to Machine Learning in the Master’s program. After my studies, I continued to work on Machine Learning at Université Paris Dauphine – PSL as a postdoc for two years. Last September, I was lucky enough to join ENSAI as an Assistant Professor.

Since then, I have been teaching how to design and analyze algorithms in the Algorithms and Complexity course. I find it to be an exciting topic because it concerns almost all programming activities.

In the Algorithms and Complexity course, I teach students how to approach computational problems, how to design algorithms (programs) to solve them, and how to analyze their performance. I also teach Object-Oriented Programming to develop reliable and maintainable software in the Object-Oriented Programming with Java course.

Both are fascinating subjects that teach us general lessons that can be applied to software design or be used in any field.

Researching multitask learning, causal effect estimation, and weakly supervised learning

Broadly speaking, I am interested in Machine Learning research. More specifically, I worked on multitask learning, causal effect estimation, and weakly supervised learning with my coauthors.

Multitask learning is a framework in which we have multiple related learning tasks. Solving them jointly by sharing information with each other can improve performance over solving independently. We wrote a paper about a multitask subspace estimation problem.

Causal effect estimation is about estimating how much a change in one variable affects another. We studied a method that can be used when these variables cannot be observed together.

Weakly supervised learning is about learning functions (input-output relationships) without directly observing the outputs, but with a weaker form of (e.g., corrupted or partially observed) analogues.

Our recent work studied a special case called Mediated Uncoupled (MU) Learning in which we only have unpaired independent observations of the input and the output variables with the help of another common variable.

The higher-level, longer-term goal of my research is to clarify when and how learners can acquire knowledge beyond that of teachers. I am greatly inspired from my teaching experience.

I like to see the ENSAI students actively learning, and I am often impressed by how they come up with great ideas that go beyond my expectations.

Mathematics as a common language

What I like most about ENSAI is the people. When I first came to Rennes to start a new life and work at the school, I was a little nervous. I had encountered so many things that I did not know how to deal with, at work and in life, but they went out of their way to help me out. I would like to take this opportunity to express my great gratitude to them. I hope I can be a part of this positive culture and become as helpful as they are.

I am really happy to be able to exchange ideas with those brilliant researchers. Sometimes, my knowledge is too limited to fully understand all discussions, but they generously and gently share their expertise. I have recently started working on a new research topic with some of them, and I am thrilled about that.

I think ENSAI offers a good balance: people work on different problems in different application domains, but we sometimes use common mathematical and engineering tools. This is, in my opinion, the best environment for building collaborations.

Communication between researchers with completely different backgrounds is useful to get new ideas but can be challenging. In our case, we can immediately start discussions using the common language. We get many new ideas in research seminars and even during coffee breaks.

How I came to teach and research

When I was still a young child, I found a science book for kids in the public library. The book showed surprising scientific facts, and I was eager to learn about science. I wanted to become a scientist who could discover such facts (if I failed to become a professional footballer).

In contrast, I did not imagine that I would be a teacher in the future. I liked explaining things to friends when I was a student, but I was not good at speaking in front of many people. I am a little surprised to find myself enjoying teaching so many students today.

It was a hard decision to make when I started my PhD program. I thought about other options. My father advised me to choose the one that I think is the most challenging, which was continuing my studies. His words encouraged me to make the decision. We’ll never know if the decision was the best or not, but I am proud of how I chose my own path.

When not a work…

I love playing football. I started playing it more than 20 years ago, and it has always been my favorite thing to do. It is also a great way to make friends with people who speak different languages.

I haven’t had the chance to play in Rennes yet, so if anyone is looking for someone to play with, please talk to me!  Another hobby of mine is exploring different beers. My current favorites are Belgian dark ale and West Coast/New England IPA. I also love cider!



2015-2019: PhD in Complexity Science and Engineering at the University of Tokyo
2019-2020: Postdoc at the University of Tokyo/RIKEN
2020-2022: Postdoc at Université Paris Dauphine
2022-present: Assistant Professor at ENSAI

Find out more about Ikko Yamane and research at ENSAI