Implementation of stochastic noise in a Hodgkin-Huxley model but as a current

Science, Projects, and Showcases
,
4

Hi all and thanks for your replies.

  1. @rth The PoissonInput would just model noise as random spikes, however, it follows a Poisson distribution and the noise would not be white. Also, it depends on the model of the synapses instead of adding a current source inside the neuron equations.

  2. @mpagkalos Yes, that would work, and thanks for the reference to the Clopath paper. However, I was looking for an implementation in a Hodgkin-Huxley model, where I do not have tau defined like so, but it’s rather more complicated. Furthermore, xi is not unitless, from the Brian2 docs it’s unit is 1/sqrt(sec).

What I was looking for is something along the lines of the following:

'''
dV/dt = (- I_L - I_noise) / Cm : volt
    I_L = g_L * (V - E_L) : amp
    noise = ... : amp
'''

What would the noise term have to be in the above to work properly as a random variable such that X~N(0,10pA)? Also, in my implementation I avoid using xi but rather randn(), as per this post.

In short, the noise as posted by Nina is given by:

'''...
noise = sigma*(2*gl/Cm)**.5*randn()/sqrt(dt) : volt/second (constant over dt)
'''

which is just voltage.

Thanks again!

1 Like