search

CINPLA / neuronify

An educational neural network app using the integrate-and-fire method.
http://ovilab.net/neuronify
GNU General Public License v3.0
54 stars 12 forks source link

Adaptive and bursting neurons #111

Open T-dogg opened 8 years ago

T-dogg commented 8 years ago
  1. Textbox should explain what parameters you have changed/added to get neurons with such behaviour. I might have overlooked something, but the basic parameters looked similar too me.
  2. With standard parameters, bursting neurons give about 20 spikes when stimulated by touch sensor. A more biological set point is 5.
dragly commented 8 years ago

Noted. Currently, bursting and adaptive neurons have an extra current that is enabled to either induce or reduce firing. As you say, the other parameters are pretty much the same as the default ones for the other neurons.

We could perhaps make the implementation details (the function forms of the currents used in the different neurons) visible from within the app somehow.

The spike count needs to be fixed to the set point you suggest. What do you suggest to be a proper firing rate / spike interval?

T-dogg commented 8 years ago

Firing rate is always a problematic parameter since it depends on the time window. I think 10 spikes within 400ms is often seen with current injections. If you consider a typical interspike-interval of 10ms you get a rate of 100Hz. Furthermore, it varies between brain areas.

Below is the defenition from a seminal paper from cortex: Regular spiking (RS) neurons were defined if the mean inter-spike interval (ISI) was over 20 ms and all inter-spike intervals (ISIs) throughout the spike train were within 15% of the mean ISI. If a ratio between the shortest ISI and the mean ISI was less than 10%, the spiking pattern was described as intrinsic bursting (IB) (Connors and Gutnick, 1990).

dragly commented 7 years ago

Let us aim for adding back bursting neurons around the time we add support for community-written plugins.