I was looking for ways to measure Flow using EEG, when I found this old neural network concept called Dynamic Recurrent Neural Networks (DRNNs). That seem like an interesting way to learn dynamics online. All of the learning rules appear to only require local information. The one way it might not work is that it seems to require changing time-constants depending on signal error, which I'm not sure how how to implement with the NEF.
I was looking for ways to measure Flow using EEG, when I found this old neural network concept called Dynamic Recurrent Neural Networks (DRNNs). That seem like an interesting way to learn dynamics online. All of the learning rules appear to only require local information. The one way it might not work is that it seems to require changing time-constants depending on signal error, which I'm not sure how how to implement with the NEF.
It's been previously used for generating acceleration data from biced and tricep recordings.