IO

896 Words2 Pages

2.2.2. IO-DCN synaptic plasticity
The MF-DCN synaptic plasticity mechanism was previously hypothesized to be a proper cerebellar gain controller which self-adapts its maximum output activity to minimize the inhibition impact of the inhibitory pathway already described (Garrido et al., 2013a). Nevertheless, this cerebellar gain controller reaches the adequate state through the learning process. This involves a time period in which the control action is not delivered properly which make the system prone to become unstable. The cerebellum, during this learning process, shall be able to supply enough control action to avoid these possible destabilization inconveniences. Furthermore, the feedback action in cerebellar motor control is indeed well accepted (Kawato and Gomi, 1992;Stroeve, 1997;Desmurget and Grafton, 2000;Kalveram et al., 2005) and there also exist neurophysiologic evidences suggesting that the primary motor cortex is involved in this feedback loop (Sergio et al., 2005). Concretely, there is a dense projection from primary motor cortex to the spinal cord, often directly onto motor neurons, and correlations between primary motor cortex activity and end-effector kinematics (Todorov, 2000). Hence, proprioceptive signals encoding for instance position error information (inputs) are put in relation with the corrective cerebellar output, thus leading one to believe that the IO-DCN connection might implement this loop.

According to Figure 1.B DCN input signals (proprioceptive signals) are received from two differentiated pathways. The first pathway reaches the DCN cells through the cerebellar cortex. This feedback system has been profusely hypothesized to be the main adaptive pathway in which cerebellar learning takes place (S...

... middle of paper ...

...ans that the weight factor is quickly decreased. As we see the potentiation/depression action compensates each other. What it is quickly learnt due to an action is quickly forgotten due to the opposite action.
In order to obtain a numerical evaluation of the Modulated Term impact in the convergence speed process (Fig 6), the normalized mean absolute error (MAE) convergence speed defined in (Luque et al., 2011b) has been used. This measurement is defined as the number of needed samples (iterations of the movement) to reach the final error average. To normalize the measurement, the cerebellar configuration without IO-DCN corrective action was conceived to be the worst possible scenario thus assigning a value of 1 to the obtained number of samples needed to reach the final error average in the absence of IO-DCN terms (i.e. the slowest possible convergence speed).

Open Document