Datashape team to work with Fujitsu
Fujitsu plans to invest €50 million in France to develop its competence in data processing and artificial intelligence. The major Japanese firm is seeking to work with Inria's researchers, starting with the Datashape team.
“In artificial intelligence and machine learning, Inria has outstanding expertise that is recognised nationally as well as internationally ,” explains Frédéric Chazal, head of the Datashape project team at the Inria Saclay – Île-de-France research centre. “There are many teams that are working on these topics, including some in areas that are innovative and original.” This is precisely the expertise that has attracted the Japanese computer giant, Fujitsu. The company, which has plans to invest €50 million over five years in France , has committed to a partnership with Inria, which this year celebrates 50 years of scientific excellence.
The partnership will start with the Datashape team. “We are still just beginning our work with Fujitsu. It's a first step. Afterwards, we'll see how we're going to continue this work together. It is also quite likely that other Inria teams will be participating.”
Topological data analysis: a promising field
The close relationship with Datashape in particular is not accidental. Indeed, in recent years, the team has become known for its specialising in a rapidly growing field: topological data analysis (TDA) . “The field is about fifteen years old, but has really developed in the last five or six years,” notes Chazal. “Our team has been a promoter and initiator in this field, especially with the development of our software, Gudhi[i]. It's a library that uses the most recent advances in TDA. Our goal is to make it the reference in the field.” Fujitsu's teams recently became interested in the development of TDA and the team's research work, used the software internally and have had promising results.
Two areas of research already identified
What will take place during the first year of the collaboration? Two research areas have already been identified around an objective: how to interpret data from connected objects. It will study data that are time dependent and consider tools for analysing them.
The researchers also plan to look into data structured in graphs and detecting anomalies. “It's a typical subject for artificial intelligence. But currently certain types of tasks and problems remain difficult. The idea is to provide tools to make advances in the field and offer new methods and solutions. We have some interesting ideas but we're still in the research area. The objective is really to generate competences so that Fujitsu can then develop effective tools.”
A “win-win” partnership
For several years, Fujitsu has been developing artificial intelligence. The IT giant brings together these leading-edge technologies in its system known as “Human Centric Al Zinrai”. Work carried out with Inria will benefit this system.
The research team is delighted with the partnership, which is an excellent way to share the results of their research. “For us, it's also a means to have access to issues and data in the industry. So it's an interesting basis for work.”
Initially, a few permanent members of the Datashape team and a small group of researchers from Fujitsu will meet on a regular basis to set milestones and launch the partnership. The ultimate objective of the two parties, however, is to establish a more significant partnership. Fujitsu has already announced its plan to develop a longer term partnership with Inria . “This is what we liked about the collaboration,” acknowledges Chazal. “They are not here to make a splash. There is a real commitment to establish something, to discuss and try to turn this into a long-term relationship. At that time, we shall certainly be signing a contract that outlines the precise contours of the partnership.”
[i] Gudhi is a library designed as a reference platform providing easily accessible software tools that are state of the art in the field of topological and geometric data analysis. The software was developed as part of a project funded by the European Research Council.