apply stochastic gradient descend logic to pinpoint overlapping ERPs

[neurodream]

gradual weight adjustments via backpropagation (or similar): follow promising paths stochastic: potential to get unstuck from local maxima

could be used to decipher non-oscillatory responses during natural stimulation (with fuzzily timestamped events at best)