A bump-attractor spiking neural network for motor learning based on Norepinephrine release

In order to examine the formation of predictive motor memories, typical behavioural motor learning experiments perturb participants reaching movements using an external force field, to which they rapidly adapt, and exhibit aftereffects when the force field is removed. While previous computational models can recreate the behavioral results, they do not account for the neural mechanisms involved. A model that also includes synaptic mechanisms could help to infer biological causes of neurodegenerative diseases through systematic changes in the model’s parameters. For this reason, we developed a bump-attractor, spiking neuron model of M1 proposing a synaptic mechanism to explain adaptation and washout based on reward-based neurotransmitter release. The developed model consists of directionally tuned neurons, shown to exist in M1, that encode the hand position through average neural firing. The force field is modeled through a simulated, external current perturbing the neural activity in the direction of the force field. During adaptation the perturbation of neural activity decreases as the biologically motivated Norepinephrine increases depending on the size of the previous trial’s error. Washout is implemented through an internal state shift in the motor-coordinate system during force field removal, followed by a de-adaptation. The simulations from our proposed computational model qualitatively account for both: adaptation and washout as seen in comparison to the behavioural data from (Nozaki et al., 2006, Nat Neuroscience). Thus, the model suggests for the first time a biologically plausible synaptic mechanism that can explain the main features of motor learning of external dynamics.      «

In order to examine the formation of predictive motor memories, typical behavioural motor learning experiments perturb participants reaching movements using an external force field, to which they rapidly adapt, and exhibit aftereffects when the force field is removed. While previous computational models can recreate the behavioral results, they do not account for the neural mechanisms involved. A model that also includes synaptic mechanisms could help to infer biological causes of neurodegenerat...     »