ChaosHunter can construct neural network formulas as well as standard mathematical formulas. Like the others, these formulas are completely open for inspection and transfer to other platforms.

Those of you who have previous experience with neural networks will appreciate the technological advances that ChaosHunter has made in this area. Network structures are completely controlled by evolution (genetic programming) to include:

1. Variables chosen from a pool
2. Inputs can be variables or math formulas containing variables, constants and time series indicators
3. Weights can be variables, constants, time series indicators, or formulas of those
4. Hidden neurons and connections controlled by evolution
5. Maximum size of the structure can be limited, but the algorithm promotes smaller, more compact structures
6. The complete formula for the nets is shown and can be transferred to Excel or computer programs
7. Evolution uses distributed processing across multiple cores and multiple computers
8. Nets are recurrent in that they can contain chaotic feedback from previous outputs
9. Time series indicators can be input as variables or evolved internally


In case your neural net building is for market trading and/or price time series:

10. Time series indicators included double as technical indicators
11. Evolution goal (fitness) can be profit instead of minimization of error
12. Instead of predictions of a target, the output of the net can be just buy/sell signals with evolved thresholds
13. The complete formulas for the nets can be transferred to popular trading platforms
14. ChaosHunter accepts almost any kind of bars as price data (minute, volume, tick, range, etc.)

In addition to the features above, neural networks can contain boolean and relational functions such as AND, OR, >, <, and IF-THEN if you desire to include them.