Bandeau image

e-Science

e-Science

Chapo
The increasingly systematic use of big data profoundly changes the practice, the techniques of work and the mechanism of inference, across all scientific disciplines: the ability to predict phenomena has become central in our analysis of scientific data. Moreover, in some areas such as genomics or neuroscience, the experimental technology revolution has opened the door to very large data sets with a limited number of observations. A key challenge is to support this change by developing a culture of data expertise within the DATAIA Institute to spread it to all scientific partners.
Contenu
Corps de texte

More and more complex data

On the Saclay Campus, we can find strong fields such as Physics and Biology, which, like all scientific disciplines, generate a large number of data through experiments. Much of the work of scientists consists in exploiting these data: store them, represent them, extract useful or important information from them, and then making modeling, statistical models, predictions, testing them, etc.

As these data become increasingly complex, not all scientists are necessarily equipped to respond to the challenges of these data, which are becoming more and more technical over time. In line with the work started by the Paris-Saclay Center for Data Science, the DATAIA Institute offers tools to facilitate the analysis of data by scientists.

Create new methods of analysis

To create new collaborations and boost data analysis, a new approach is to support the creation of challenges to identify, for a given problem, the best strategies and combine them to arrive at a more efficient overall model.

For example, in the framework of the Center for Data Science, researchers and engineers from the Parietal team of the Inria Saclay Center - Île-de-France, launched in 2018 a challenge on autism from images of the brain of children, autistic or not. Thus, learning specialists are developing predictive models for the diagnosis of the disease. This challenge made it possible to advance the prediction score and boost the quality of the diagnosis of autism by 10%.

In e-science, another type of approach is to establish the causal structure of phenomena rather than prediction: it is a very active branch of learning.

The DATAIA Institute facilitates interdisciplinary collaborations to develop new approaches to solve various Machine-learning problems (missing data, causal analysis, statistical control, etc.).

Go beyond

Contenus liés
Bandeau image
fleche