www.neuron.yale.edu

On 3/10/2005 Kevin Crisp <kevin_michael_crisp@yahoo.com> wrote:
At the Neuron training session at the University of Minnesota several years ago, either yourself or your colleague showed me an interesting demonstration using Neuron. It involved getting a three cell reciprocally connected integrate-and-fire model to oscillate bursts in three phases by adjusting the noise level in the circuit. I believe a reference was made to Otto Friesen's work (UVA).

I am wondering if you can direct me toward the original source (published) of this demonstration.

On 3/10/2005 Ted Carnevale <ted.carnevale@yale.edu> wrote:
> At the Neuron training session at the University of Minnesota
> several years ago, either yourself or your colleague showed
> me an interesting demonstration using Neuron. It involved getting
> a three cell reciprocally connected integrate-and-fire model to
> oscillate bursts in three phases by adjusting the noise level in the circuit.

The role of stochastic inputs is to guarantee symmetry
breaking. Otherwise the cells would merely fire tonically.

> I believe a reference was made to Otto Friesen's work (UVA).

Go to ModelDB
http://senselab.med.yale.edu/senselab/modeldb/
and search for the model by Friesen.

In retrospect, I see that a few additional comments are necessary.

> It involved getting a three cell reciprocally connected integrate-and-fire model

The connections in this net are not reciprocal. They are cyclical.
That is, A inhibits B which inhibits C which inhibits A, and A B & C
are all driven by excitatory synapses that are activated by afferent
spike events that have the same mean frequency but arrive at
independent random times.

If C speeds up, A slows down. This lets B speed up, which slows
C. That allows A to rev up, which slows B, and the cycle repeats.
So the wave of increased spiking precesses widdershins ("backwards",
or perhaps more properly, counterclockwise), i.e. the sequence is
C B A C B A C . . .

> to oscillate

The wave of increased spiking does not have a regular period,
since it depends on chance fluctuations in excitatory drive.

> bursts in three phases by adjusting the noise level in the circuit.

The emergence of a cyclic firing pattern depends on many factors:
time couse and magnitude of unitary synaptic inhibition and excitation,
conduction delays, time constant and firing threshold of the cells, and
the statistical properties of the afferent excitatory spike trains.