Initial deposit size - what is it? - page 17 1...101112131415161718 New comment [Deleted] 2008.10.24 04:12 #161 StatBars писал(а) >> I don't know why your EA is crashing, there is no code). It doesn't look like Kohonen Maps either...(at all) So on the previous page it's all there! Artem Titarenko 2008.10.24 04:58 #162 On the previous page is a network skeleton, even more so without input data. By feeding the MACD you can get values other than 0. In theory, the network should function normally if you have trained it, generated it, and now supply the same values as you did when you trained it. Cloth has a lot of NS codes on the forum and on the web, including Kohonen maps, you'd better take them... Machine learning in trading: Machine learning for robots Discussion of article "Neural [Deleted] 2008.10.24 05:07 #163 What's that on page 15? It's waiting for you. :))) Files: swettenea_1.zip 91 kb Artem Titarenko 2008.10.24 05:32 #164 Swetten писал(а) >> What's that on page 15? It's waiting for you. :))) Before MathExp(), the array feature2[] produces large values. After MathExp() these values become very small. Why this code does not work I do not know. [Deleted] 2008.10.24 05:35 #165 StatBars писал(а) >> In array feature2[] before MathExp() you get big values. After MathExp() these values become very small. Why this code does not work, I don't know. What is interesting - this grid cannot be opened in NSH2 as well. Overflow and hello. [Deleted] 2008.10.24 05:36 #166 StatBars писал(а) >> Cloth has plenty of NS codes on the forum and on the web, including Kohonen maps, you'd better get them... Can I get a link to one that works best? Or the one that's good? Or not very hard to code? Artem Titarenko 2008.10.24 05:46 #167 Swetten писал(а) >> What's interesting -- this grid doesn't open in NSH2 either. Overflow and hello... NSH2 has only seen a glimpse... Artem Titarenko 2008.10.24 05:52 #168 Swetten писал(а) >> Can I have a link to one that works best? Or a perspecƟve one? Or one that is not too complicated in terms of code? There are examples ofa simple network in MQL4 here. But you have to deal with both the training method and sampling at the input. Check here. [Deleted] 2008.10.28 21:14 #169 Guys, thank you so much for the advice, help, links and just understanding! Here's another creation. Any critique? And I'll have a sit-down meal with cake and tea! :))) Files: swetten.mq4 68 kb [Deleted] 2008.10.29 18:17 #170 In case anyone is interested, the Expert Advisor can be found in the Code Base here. With pictures! :))) 1...101112131415161718 New comment
I don't know why your EA is crashing, there is no code).
It doesn't look like Kohonen Maps either...(at all)
So on the previous page it's all there!
On the previous page is a network skeleton, even more so without input data.
By feeding the MACD you can get values other than 0.
In theory, the network should function normally if you have trained it, generated it, and now supply the same values as you did when you trained it.
Cloth has a lot of NS codes on the forum and on the web, including Kohonen maps, you'd better take them...
What's that on page 15? It's waiting for you. :)))
What's that on page 15? It's waiting for you. :)))
Before MathExp(), the array feature2[] produces large values. After MathExp() these values become very small.
Why this code does not work I do not know.
In array feature2[] before MathExp() you get big values. After MathExp() these values become very small.
Why this code does not work, I don't know.
What is interesting - this grid cannot be opened in NSH2 as well. Overflow and hello.
Cloth has plenty of NS codes on the forum and on the web, including Kohonen maps, you'd better get them...
Can I get a link to one that works best? Or the one that's good? Or not very hard to code?
What's interesting -- this grid doesn't open in NSH2 either. Overflow and hello...
NSH2 has only seen a glimpse...
Can I have a link to one that works best? Or a perspecƟve one? Or one that is not too complicated in terms of code?
There are examples ofa simple network in MQL4 here.
But you have to deal with both the training method and sampling at the input.
Check here.
Guys, thank you so much for the advice, help, links and just understanding!
Here's another creation. Any critique? And I'll have a sit-down meal with cake and tea! :)))
In case anyone is interested, the Expert Advisor can be found in the Code Base here.
With pictures! :)))