Please check this blog post. This is about the Spinal Implant EA.
What we do with Spinal Implant and the perceptrones is that we use the native mt4 genetic algorythm.
1. We make a classification: The neural net provide output signals. The general interpretation of these signals will be some sort of "binary" classification, like "buy" (>0) or "sell" (< 0).
2. This neural net is "unsupervised" neural network, and that is meaning that it does not need to be trained by showing the correct answers, like most neural networks (i.e., "supervised" neural networks). You do not teach them by providing any kind of actual output which they learn to reproduce, as with most neural nets.
What really need to be understand by everyone who will download those codes, is that the percetrone and kernel codes do not predict anything.
What we do is to use the genetic optimizer who will evolve the weights leading to the best results.
The Neural net "learns" how to give its signals based upon evolutionary pressure provided in the native MT4 algorythm. The genetic algorithm (GA) in the MT4 "evolves" the neural net that give better and better signals. Survival of the fittest controls the evolutionary process as usual, where fitness is determined by how much money the NN makes (you can use what to optimize in the MT4, e.g. Balance, Profit Factor etc.).
When you use other software the first thing is to export the data, to build a neural net model and then import the neural net model into mt4.
There are many tutorials and different ways to do it. As software there is indeed a choice.
For example arryex from trade to win use Neuroshell to build the models.
There are examples with Neuro solutions
There are examples with of course Matlab.
However one of the most interesting things is the open source implementation with the JNSS. The JNNS is a german neural net simulator, that gives many different neural net architectures.
I was thinking to use the export and import codes in order to build custom EAs, but as the native MT4 algorythm together with our native codes exceeded really my expectations I never did that.
Comments
take a look also on this :
http://www.dtreg.com/mlfn.htm
Cascade correlation for tuning perceptrons
http://www.ra.cs.uni-tuebingen.de/SNNS/UserManual/node164.html
http://www.lri.fr/~antoine/Courses/DEA-I3/Tr-IA-inapg-NN.pdf
Please check this blog post. This is about the Spinal Implant EA.
What we do with Spinal Implant and the perceptrones is that we use the native mt4 genetic algorythm.
1. We make a classification: The neural net provide output signals. The general interpretation of these signals will be some sort of "binary" classification, like "buy" (>0) or "sell" (< 0).
2. This neural net is "unsupervised" neural network, and that is meaning that it does not need to be trained by showing the correct answers, like most neural networks (i.e., "supervised" neural networks). You do not teach them by providing any kind of actual output which they learn to reproduce, as with most neural nets.
What really need to be understand by everyone who will download those codes, is that the percetrone and kernel codes do not predict anything.
What we do is to use the genetic optimizer who will evolve the weights leading to the best results.
The Neural net "learns" how to give its signals based upon evolutionary pressure provided in the native MT4 algorythm. The genetic algorithm (GA) in the MT4 "evolves" the neural net that give better and better signals. Survival of the fittest controls the evolutionary process as usual, where fitness is determined by how much money the NN makes (you can use what to optimize in the MT4, e.g. Balance, Profit Factor etc.).
When you use other software the first thing is to export the data, to build a neural net model and then import the neural net model into mt4.
There are many tutorials and different ways to do it. As software there is indeed a choice.
For example arryex from trade to win use Neuroshell to build the models.
There are examples with Neuro solutions
There are examples with of course Matlab.
However one of the most interesting things is the open source implementation with the JNSS. The JNNS is a german neural net simulator, that gives many different neural net architectures.
I was thinking to use the export and import codes in order to build custom EAs, but as the native MT4 algorythm together with our native codes exceeded really my expectations I never did that.
I was thinking about something else when writing this blog
- to try to find the best combinaison of weights to fetch the best correlation to the price
Means,
a series of 4 weights during 300 bars, calculate the sum of the each correlation values to the price, until the best combinaison is found.
this is totally different about the binary classification.
Yep, I need now a way to calculate w/ C language and send the result to Metatrader 4