Discussion of article "Neural Networks: From Theory to Practice" - page 8

[sigma7i]

Great article!!!
Only the question of an example of a system of 2 or more neurons remains open.

Maybe someone has a simple example of a 2, 3 neuron advisor?

[Dmitriy Parfenovich]

sigma7i:

Great article!!!
Only the question of an example of a system of 2 or more neurons remains open.

Maybe someone has a simple example of a 2, 3 neuron advisor?

I was planning to write a second article on this issue, but I don't have time yet.

Try this design

double out_net; //network result
double out[2];  //results of the first layer of neurons

out[0]=CalculateNeuron(inputs0,weight0); //computing the first neuron in the first layer
out[1]=CalculateNeuron(inputs1,weight1); //computing the second neuron in the first layer
out_net=CalculateNeuron(out,weight2);    //computing the output neuron of the network

I think the idea is clear.

[sigma7i]

fyords:

Try this design.

I think you get the idea.

It's very clear, thank you!

Here it turns out that the number of weight coefficients doubles with each neuron, but it's a matter of creativity :).

I'll wait for the article!

[TipMyPip]

MetaQuotes:

New article Neural Networks: From Theory to Practice is published:

Author: Дмитрий

Dear Sirs, Thank you very much for the article, will it be possible for you to change the (,) commas into floating points?

 

733,56

2 658,29

 

733.56

2 658.29

[Paul Anscombe]

This is a great article well done.

I have some questions though...

In your 1 neuron example the input is the last 10 period values of the RSI indicator. Therefore, the output for that neuron is simply going to be a sophisticated form of weighted average for the last 10 RSI values, is this how you would envisage indicator data being used in reality?

For instance if I wanted to use 3 indicators as inputs, would you expect to implement 3 neurons in the manner in your article cascading to a 2nd layer neuron, or would you simply use the last value of each of the 3 indicators as input into a single neuron?

 

My other question is in a multiple layered network, would you still need to normalize the data from the first layer to input into the second layer, given that it will be in the range -1,1  or 0,1 anyway?

 

many thanks

 

 

 

  

[Jin]

Did anybody try to emulate the results?

All my tries result in downward balance in Forward results.

Also, the number of ticks processed doesn`t match - it`s almost half of what is indicated in the picture.

One thing that took my attention: the number of ticks 17331 from the period between 2012.01.02 and 2012.09.14 matches exactely if I disable the Forward option. Hummm ...

[enbo lu]

Good article which make it easy to understand the basic concept of neural networks. It help me a lot. Thank you!

[[Deleted]]

Thank you.
An interesting example, which uses the concept.   https://www.mql5.com/en/code/1649

[surubabs Suresh Kakkattil]

MetaQuotes:

New article Neural Networks: From Theory to Practice is published:

Author: Дмитрий

The best example on the base of article is https://www.mql5.com/en/code/1649 ,

Most traders who use Bolinger Bands, are serching for Bollinger Width based EA,

The EA which i have posted give Width of Bollinger band, it dosent use any iCustom indicators, all the calulation has been done on the base of Bollinger band indicator,

With the use of Neural Network method, you can see that the width EA trading,when its break out.

Realy interesting it is watch it yourself.  

[Cagatay Tezsezen]

Great article. However, this method takes you to a result, curve fitting. Eventually, real account live tests may be disapointment. The information given in this article is valuable for those who wants to understand how to  start in AI, but they have to find a better way of implementation to avoid themselves from curve fitting methods. Any trading system must be able to use some dynamic parameters which will reflect the changing conditions of market. Otherwise, EA will be out of date in a very short time leading your account to 0 balance. So, the input has to be designed very carefully. The important thing is the design of inputs and you need to know what the output shall look like. AI does not mean, you will give any input and get some great output.  AI developers have a very common word, "GARBAGE IN, GARBAGE OUT"