Section: New Results

Machine Learning

In Machine Learning, we were interested this year in two phenomena for which we consider classical paradigms of modeling and for which we wonder how they could be adapted by bio-inspiration.

The first paradigm concerns the manipulation of temporal sequences. In a perspective of better understanding how brain learn structured sequences we extended a model on syntax acquisition using the Reservoir Computing framework (using random recurrent networks) [16], [9], [19], [20]. The extended model is also used in a Human-Robot Interaction architecture to enable users to use more natural language with robots [14], [15], [18]. This work will be extended with our collaborators at the University of Hamburg (cf. § 9.3).

In an industrial application for the representation of electrical diagrams (cf. § 8.1), we also study how recurrent layered models can be trained to run through these schemes for prediction and sequence representation tasks.

The second paradigm concerns the extraction of characteristics and the use of hierarchical networks, as in the case of deep networks. An industrial application whose study has just begun (cf. § 9.2) will lead us to revisit these models to make them more easily usable in constrained frameworks, for example with limited size corpuses.