Fast Neural Plug-in

Plug-in Description

Figure 1. The nodes available for the fast neural net plug-in module.

This neural plug-in module performs simulations that are more biologically realistic than standard connectionist schemes, but it is still abstract enough to be able to run much more quickly than more detailed integrate-and-fire or compartmental models like Neuron or Genesis. It is a firing rate model based on work originally done by Beer. The neuron is modeled as a simple RC circuit that is integrated using the Euler method. The firing frequency of each neuron ranges from 0 to its normalized maximum of 1. Synapses between neurons have a weight that is the maximum amount of current that will be injected into the post-synaptic neuron which is proportional to the firing frequency of the pre-synaptic neuron. Modulatory and gated synapse are also provided to allow the firing of one neuron to alter the gain of the output of another neuron. No learning is currently implemented in this model.

The following pages provide more details about how this neural model is implemented, and the types of neurons and synapses that are available. Also, the global properties for this module are listed in the next section.

1.	Neuron Basics.
2.	Neuron Model.
3.	Numerical Integration.
3.	Normal Neuron.
4.	Pacemaker Neuron.
5.	Random Neuron.
6.	Tonic Neuron.
7.	Bistable Neuron.
8.	Synapse Types.

Plug-in Properties

You can find the global properties for the plug-in modules by clicking on the modules tab of the behavioral editor window. The properties dialog box from figure 1 will be displayed. To edit the properties for the fast neural plug-in click on the FastNeuralNet listing in the neural modules list. This will display the following properties for you to edit.

Network Filename
The name of the file for this neural plug-in module for this organism. This is a read-only property.

Figure 2. Fast Neural Net Plug-in Properties

Time Step
This is the time step to use when simulating the neurons in this plug-in module. Each plug-in can have its own time step defined. This allows you to run simulations with different levels of detail. For example, since the neurons in this model are more detailed you may want to run them with a time step of 0.2 ms. But a less advanced model like the Fast neural net plug-in may only need to run at 2.5 ms. This lets you run the less detailed model at a much faster rate. The simulator will compare the time steps for all the modules and the physics engine and determine the lowest time step. It will run all the other modules using some integer modulus of that minimum step.
Default value: 2.5 ms.
Acceptable range:: Greater than 0.