fzenke / auryn

Auryn: A fast simulator for spiking neural networks with synaptic plasticity
https://fzenke.net/auryn/
GNU General Public License v3.0
98 stars 42 forks source link

Added refractory period functionality to AdEx. #22

Closed sanjayankur31 closed 9 years ago

sanjayankur31 commented 9 years ago

Even though AdEx has the capability to adapt, in cases where the input current or excitatory conductance is very high, for example in case of a network of excitatory neurons, the firing rates of the AdEx can be very very high - which is not biologically plausible - since the exponential term comes in to play too quickly. To remedy this, a refractory period must be included.

I'm leaving the default refractory period as 0, which is closest to the original model, but I've added the functionality required to set a refractory period if users require it.

The following figures will show the differences.

Without a refractory period: adex-0ms-refractory

With a 2ms refractory period: adex-2ms-refractory

The F-I curve for AdEx without a refractory period - notice the very very high firing rates that the model is capable of in comparison to the TIF with a 2ms refractory period - clearly not biologically plausible. The upper limit here depends on the integration time step in use: adex-ab0-tif-fi