Neural Networks

Hi, I’m new to neuro science and networks, I’m trying to figure out why is it that exciting an ‘inhibited’ neuron has more effect than exciting an ‘excited’ neuron? What could be the reason for this?

Thanks very much for your help in advance!


Let me start by saying that I think what you wrote isn’t necessarily generalizable across all neural nets; there are plenty of parameters that would determine which has more of an “effect”, and this also depends on what scale we define “effect” (e.g. is it local / from that neuron, or global / whole network dynamics).

To most directly answer what you asked, consider the sigmoidal function, which is a common activation function in neural nets:

Let’s say at a given point in your network simulation that a neuron is “excited” at z=5. If you further excite it, let’s say to z=6, look at the change in phi(z); you should see that it is very small, thus the output from that neuron will only change marginally. Now let’s excite an “inhibited” neuron from z = -4 to -3. You should see that the y-axis change is more dramatic here, and thus so is the change in neuron output.

This is all at the local level, that is, just considering that neuron itself. How that affects the global network dynamic depends on much more parameters that, such as the weight of that neuron’s output and how influential its neighbors are.

Hoped that helps!

1 Like

Nah mate, your answer does not help me. It looks like you did not understand what I meant. Thanks for putting in the effort for trying, but that does not answer my question.


Hi Poyraz,

Can you try to rephrase your question?


Hi Steven, it is too late for that now, no answer will help from now onwards. :disappointed: