Disliked{quote} I am currently studying possible use of Neural Networks using MACD indicator to find a dominant pattern automatically. I found one simple way to do data scaling of that indicator. Basically I tried to divide indicator values coming from JPY by 100 to get the right scale compared to other pairsIgnored
- One may suggest that all patterns are Waves (1-2-3-4-5-A-B-C) or harmonic patterns. But I advocate differently, I say they are simply all M patterns in different fonts which makes it only logical that a 'font recognition robot' (not a trading robot, but a program that reads handwritten input from academic thesis papers by using Average-Pooling or similar Deep Learning techniques) adds as much value to it as, what you use to calculate (dividing 100). Have you ever used a font recognition program? If yes, do you see the effectiveness if used on the GBP/JPY or EUR/USD chart?
- Regarding your reply on dividing by 100. Okay, are you comparing all (major) currencies with the JPY? You mention 'dividing' in your formula, where I can only assume you would add a multiplication to your calculation as well. What do you multiply? Could you explain more about your formula as to recognize a pattern (in the MacD)?
- Neural networks, a concept where you add lotsize to probability aspects, am I summing it up correctly? Are you willing to list the different values (probability vectors and dropouts to avoid overfitting) in your concept (HWL) of pattern recognition with neural networks?
Join our skype group.
