Computational Embodied Neuroscience ( 24 hours)
Teachers: Gianluca Baldassarre (CNR), Daniele Caligiore (CNR), Emilio Cartoni (CNR)
The course aims to teach students how to build computational models of behaviour and the underlying brain mechanisms. A key features of the course is that the models will be `system-level models’: this means that they will focus on how behaviour is produced by a system of integrated brain components working together rather than by a single one (e.g.: multiple parts of basal-ganglia, cortex and cerebellum). When possible, the models will also be `embodied’, meaning that they will be tested within a simulated body interacting with a simulated environment similar to real organisms. The computational models will be implemented in Python. The specific topics of the course will be as follows:
– Why computational models to study the brain: neural mechanisms and adaptive functions
– System computational neuroscience: brain as a complex adaptive system and behaviour as an emergent phenomenon
– The central role of learning and motivations for the study of motor/cognitive development and brain diseases
– Models of neurons: firing rate, leaky units, spiking
– Neural-network architectures: feed-forward, recurrent, bio-constrained
– Modelling cortical perception with unsupervised learning mechanisms
– Modelling cerebellum forward/inverse models for motor control with supervised learning mechanisms
– Modelling motor learning (habitual behaviour) with reinforcement learning mechanisms
– Modelling goal-directed behaviour with firing rate and spiking models
– Anastasio, T. J. (2010). Tutorial on neural systems modelling. Sunderland, MA: Sinauer Associated.
– Trappenberg T.P. (2010). Fundamentals of computational neuroscience. Oxford: Oxford University Press.
– Rolls E. T., Treves A. (1998). Neural networks and brain function. Oxford: Oxford Unversity Press