Section: New Results

On the computational efficiency of Basal Ganglia models

Many valuable models have been proposed to capture the richness of the fundamental relations between the basal ganglia and prominent brain structures including the prefrontal cortex, the hippocampus and the superior colliculus. To choose among them the mechanisms on which to build the design of the motor pole of our brain-inspired system, a fundamental issue is to evaluate the efficiency of these models in more realistic cases than the ones which are generally considered by the authors [24] . For this reason, we have conducted a systematic study of several basal ganglia computational models to check of their scalability in terms of ation representational space [25] . Unfortunately, we found most of them to not be scalable and some of them to not be reproducible at all.

Another way to explore the computational efficiency of neuronal models is to implement them at lower levels of description. This is currently being done with one model developed in our lab at a level corresponding to a neuronal assembly with a mean activity expressed using a single variable. This mesoscopic approach has been refined to a microscopic scale description level, i.e taking into account individual neurons and synapses. Besides the confirmation of many of the results of initial model with a more detailed formalism, this new model has allowed us to highlight the facilitating role of inhibitory interneurons in the decision-making and action selection processes.