Project idea – oscillators and clocks (adaptive or not here we come!)

It was mentioned that oscillators and clocks play a crucial role in cognitive systems. While it is rather simple to write computer code that times something, or using a sin or cos function to create an oscillation for Markov Brains we have no idea how they would do that (there are publications for ANNs). But what seams to be more interesting is the synchronization of rhythms. Imagine I start clapping in a certain frequency, you will have no problem to pick up and clap at the same time I clap. Fireflies for example do that too, and we have experiments done with AVIDA how they synchronize with each other, but not to an external signal.

So what I would suggest is the following experiment:

Feed a regular sequence of zeros and ones into a Markov brain (or ANN), where the ones are separated by a fixed number of zeroes (like: 10001000100010001000 …), we should now try to evolve the MB or ANN to always signal a 1 at the same time when a 1 gets fed in. It is important to note that the network should not produce a 1 after the 1 was fed in, but at the same time (or maybe at t-1). It will be more than trivial to evolve systems that have a particular rhythm, what becomes interesting are changes to the sequence that should be “learned” during the lifetime of the machine. For example I repeat 1000 for a couple of times and then change and repeat 100. The brains should pick up this change and adapt accordingly. In order to do so brains must of course be evolved in environments that reward this synchronization behavior.

I am not sure, but I think a lot is known about actual neural processes that create these oscillations and rhythms, it should be very interesting to compare the evolved solutions to the actual biological systems. At the same time, we would have evolved a systems that is capable of lifetime adaptations, which itself is rather cool in my opinion.

 

3 Comments

  1. Peter March 12, 2014 10:23 pm Reply

    This isn’t really my area of psychology, but ones of the labs here does a lot of work on time perception. Apparently people are very good at synchronizing to a beat, and can usually continue to generate it at the same frequency for a while after the external beat stops. But there are some funny constraints where the beat has to be within a certain frequency (e.g. 1 / min is too slow), and people’s ability to keep the beat varies with age (little kids do much better with faster beats) and ability (people who stutter are pretty bad at this).

    I think they’ve found that people also have relatively narrow windows for preferred beat frequencies (when you ask them to spontaneously produce a beat), implying that there’s probably some range of oscillator frequencies that people have access to. So they might be able to do 1000100010001000 but not 1000000001000000001 or something to that effect. I’m not sure what implications this has for evolved MB that can do this task…

  2. Hintze March 13, 2014 1:43 pm Reply

    We have a couple of project dealing with the evolution of synchronisation for mating or predation protection in fire flies, and they also have a narrow range of rhythms they can do.

  3. Jory Schossau March 13, 2014 1:47 pm Reply

    I wonder about detection of a simple pattern vs if it’s seeing noise, but maybe that’s a more complex task than the one proposed.

Leave a Reply