The context of the project is the study of networks of interacting nonlinear systems with application to neuroscience. The project adresses two timely and innovative questions: - Are we able to estimate the connectivity/directivity graph from the observation? - What is the influence of the network topology on the information processing of the network? To answer the questions we develop four tracks of research: 1- Practical inference for connectivity and directivity in complex networks (statistics, kernel methods, information theory) 2- Synthesis of multivariate signals interacting according to a complex network. (stochastic processes) 3- Information processing by pooling networks (information theory, modeling of neural of spiking neurons) 4- Taking into account directionality : directed information theory in neural networks (feedback/feedforward in neural networks) Outgoing phase of the project hosted by the University of Melbourne, Dept. of Mathematics and Statistics with collaboration with the neuroengineering group and the Center for Neural engineering. Return phase hosted by CNRS/ GIPSAlab, Grenoble, France. Expected outcomes and benefits for the EU includes: 1- The design of a practical methodology for inference of connectivity/directionality between signals with nonproperties (nonstationarity, nonGaussianity, ...). Designed for neuroscience applications, but easily extended to other scientific domains. 2. Enriched understanding of the influence of the topology of networks onto information processing. 3. Enriched understanding of the importance of feedback/feedforward onto the information processing of neural networks. Anticipated related outcome includes improvements in prostheses, brain computer interfaces, better understanding in diseases such as epilepsy. Expected experience acquired in management of mulidisciplinary project through the collaboration with Math/stat dept and Center for Neural Engineering