Right now I’ve been toying around with Adaptive Exponential (AdEx) spiking neurons in Python 3.5, since (at the moment, at least) I feel that spiking neural networks might give more realistic behavior and “thought”.
I keep running into issues with SNNs, though, the most recent being this: the exp() function. So, the first couple of questions: How do I utilize exp() without it spouting overflows and “nan”s? Are there any alternatives? I am using the equations from a wiki that look like this:
C dV/dt = −gL (V − EL) + gL ΔT exp(V − VT ΔT) − w + I
τw dw/dt = a (V − EL) − w
And the final question:
Is it true that AdEx neural networks really require negative values as their parameters?
I can post the code if necessary. Thank you for any help you can give me.
Edit: I put “closed” instead of “solved” because I’m now experimenting with alternate spiking neurons. Thanks.