Mesoscope Equipe

The sudden availability of multidimensional datasets, and the recent developments in machine learning, have opened a whole field of issues and opportunities. Clustering, text-mining, event and anomaly detection are some of the analysis that allow to extract meaningful structures from otherwise unstructured datasets. Nonetheless, the interpretation of results derived from such algorithms are often abstract and inherently difficult to communicate and visualize. In this proposal we aim to experiment and develop exploratory visual interfaces designed to visualize and interact with multidimensional datasets and more specifically with structures and components extracted from them. These challenges will be tackled both from the scientific perspective ‐ where suitable analytics will have to be devised ‐ and from a design perspective, with the goal of exposing information in insightful and inspiring ways. Methodologically, the team will work on a preliminary analysis of the different typologies of structures, discuss possible general approaches to the visualization and interaction with structures of multidimensional data, devise and prototype proof of concepts for promising directions. While a theoretical approach may be necessary for a definition of the object of experimentation, the overall aim is to allow for a workshop-like environment characterized by hands-on experimentation and prototyping based on real-world data and real-world exploration needs.

The project plans to involve the expertise of the complex systems team at the CPT Marseille (especially with regard to network-data structures). CPT Marseille has ongoing collaborations with members of the team with regard to research on dynamical networks and has collaborated on the development of one of the datasets the team will work on. The team will also take advantage of the residency to collaborate with the artists and scientists in residence at IMéRA.