Tal Neiman and Yonatan Loewenstein, Covariance-based synaptic plasticity in an attractor network model accounts for fast adaptation in free operant learning ,
The Journal of Neuroscience, 01/2013, Volume 33, Issue 4, p.1521-1534
Supplemental animations:
- Animation 1.
A simulation of the network dynamics (Eqs. 1 and 2) with $g_1=g_2=0$. A. the
activities of the two populations over time. B. the location of the corresponding
particle in the double-well approximation (Eq. 14). The tail of the particle depicts its
location in the preceding 250 msec. Note that the state transition at time $t=1.23$, marked by
a circle and a by change in background color, is in slow motion. The magnitude of
noise in the simulation is $\sigma=0.3095$.
- Animation 2.
A simulation of the network dynamics with synaptic plasticity (Eqs. 1-4). A. the activities of the two populations over time. B. the external inputs $g_1$ and $g_2$. Times of rewards in target 1 and 2 are marked by blue and red dots, respectively.
C. the double-well approximation. Note that each reward changes the shape of the
double-well potential. Parameters used: $\phi=0.016$, $\sigma=0.3095$. Reward schedule is concurrent VI with baiting rates ratio of 1:1.
- Animation 3.
A simulation of the dynamics of the model (Eqs. 1-4) in a concurrent VI schedule in response to a change of the baiting rates ratio from 9:1 to 1:9 at time $t=100$ sec (marked by a vertical dashed line in A-C). A,B and D, same as in Animation 2. C. the running averages of the population activities (Eq. 4). Note that the time for a full adaptation is approximately 150 sec. Parameters used: $\phi=0.016$, $\sigma=0.3095$.