hi, can someone explain how neural network is applied in (systematic) trading in very very simple layman terms? thanks,
Without getting into details, past returns are used to train a neural network to estimate future returns, N days ahead. This is basically curve-fitting and one reason the popularity of this method fainted after it was tried a few decades ago. If market conditions remain the same the NN prediction will be fine but when they change it fails and the long term performance is near 0 minus trading cost.
It's a mathematical construct to fit a model to historical data, and use this model to forecast the future.
NN's find patterns (regularities) in the data, and try to use those patterns to predict a label (outcome). these patterns don't have to be exactly the same, they can be just similar (this is why they are good at handwriting recognition), but the labels have to be consistent. the tricky part when it comes to finance, is setting the labels, it is hard to get a clear definition of what the labels should be, what you're trying to predict and whether or not you actually predicted it.
nonlinear5 Good explanation. So one finds a pattern that fits ones theory Watched a webinar once, the guy had to go back to 1985 to prove his point.