www.neuron.yale.edu

Hey, I had a question concerning some behavior with some code I wrote to make ring-shaped neural networks with different numbers of neurons. Just as you would expect (and like the example in Translating network models to parallel hardware in NEURON), each neuron has two synapses and NC - one to each of its nearest neighbors. So, the thing is, I am getting different results as the network gets bigger. Below, I posted images of the potential over time in 3 different simulations. Each simulation injected an identical current into one of the identical neurons in the network, and the only difference in each network is the number of neurons in the ring.

For a ring with 5 neurons:



10 neurons:



25 neurons:



As you can see, while the behavior is pretty much the same between a network with 5 and 10 neurons, where the spike just sort of oscillates around the network, with each neuron activating the next, by the time I got to a network with 25 neurons, the same current injection didn't end up exciting any neurons except the one that the current was directly injected into. Any guess why this behavior is cropping up? Since each neuron in my model is identical, I don't know why the size of the overall network should matter; wouldn't the fact that each neuron is only connected to two identical neurons mean the same behavior should occur in all 3 models, but that the amount of time between spikes for the same neuron should be larger as the network grows larger?

There aren't any figures, but that doesn't affect the following answer because the default hypothesis is always "there is a mismatch between the conceptual model and the actual implementation." So: have you verified that the properties of the model cells and synapses are as you expected, and that the the connectivity, weights, and latencies of the connections are what you think they should be?