NOTE: In NeuroShell Trader, open the chart named “Example 15 – Optimize Across All Chart Pages” which is the basis for following example:
This example is just like the last one (Example 14 ) except for two things:
1. Instead of each stock (chart page) being a separate optimization, we performed the optimization across all chart pages. That way, the neural net inputs and thresholds will be the same for all stocks. This is another way to help reduce over-fitting by building a more “general” model. However, it comes with a price – only XLE performed well in the paper trading period in this case.
2. We removed the specification to optimize for exactly 3 minutes because, with more work to do, we wanted the optimizer to run until it stopped making progress for several generations.
Look in the tab called “Modify Prediction Parameters” ->”Optimization” to see how we specified the operations above.