Habituation of sensor values through synaptic filtering

Hi everybody,

I’m wondering if anyone knows of any example of using spiking neural networks for habituation with sensors/sensor systems? In other words, the idea that when a sensor’s signal or signal statistics stay the same for a while, the response drops off.

I’ve read about this phenomenon in some neuroscience literature that discussed synaptic filtering, like this review paper (see under “Adaptation and enhancement of transients” but I’m just curious about if anyone actually tried something like it in an “engineering” context. Seems like there could be some interesting applications. Thanks!

Hello!

It sounds like you’re describing a high-pass filter. I’m not aware of any work that’s been done toward implementing that system, but there’s possibly relevant work looking at computing derivatives you may find interesting!

Yeah, it’s a high-pass filter, but I guess what I was thinking about was that in the synapses this filter wouldn’t really be “static” (i.e. cutoff frequencies and gains wouldn’t necessarily stay the same).

From the same article:

“The filtering characteristics of a given synapse are not fixed; they can be adjusted through modulation of the initial release probability or other aspects of synaptic transmission”

I’ll check out the paper you linked, thanks! And I’m guessing something like this shouldn’t be too hard to implement in Nengo