Whether dynamic or static, discrete or continuous, a neural network is a probability density function. Back-propagation is a means of determining weights in a static feed-forward neural network. GE employed them in clothes dryers for automatic shut-off back in the 90's, essentially replacing Kalman filters.