Buesing, Bill, Nessler, Maass.
Neural Dynamics as Sampling: A Model for Stochastic Computation in Recurrent Networks of Spiking Neurons.
PLoS Comput Biol, 2011.
The default setup is comparable to Figure 3 of the above paper and plots some statistical evaluation of simulation results. Three different network dynamics are supported:
(1) Absolute refractory mechanism w/ rectangular PSPs,
(2) Relative refractory mechanism w/ rectangular PSPs,
(3) Relative refractory mechanism w/ additive, alpha-shaped PSPs.
For network dynamics with absolute refractory mechanism and rectangular PSPs, convergence to the
target distribution is guaranteed for T --> infinity. The other classes exhibit more
biologically realistic dynamics, but only provide approximate samples from the target distribution.
Please refer to the paper for further details.
In case of questions, please contact
Lars Buesing (lars [at] gatsby.ucl.ac.uk) for general questions
Johannes Bill (bill [at] igi.tugraz.at) for software related questions