"New Neural" is an Open Source neural network engine project for the MetaTrader 5 platform. - page 88

[fellow]

theory continues

[fellow]

I'll have to make one myself, even though I don't quite understand these networks, but I have no choice

[nowi]

And I propose to create a cybernetic mechanism like the Terminator with a neural network encoded in the chip and send it to the future to read quotes and send them to us at the output

[nowi]

By the way, there was already a similar project when a group of talented programmers created a very smart neural network for forex and it traded a lot of money.

But the network was so clever that it withdrew all the money from the account and disappeared in an unknown direction. No one has heard of it since.

[Rashit Ayupov]

Urain:

PRNG class in the specified range.

Advantages: possibility to have several randoms with different initialization,

evenly distributed values of any range up to 10 lm (unfortunately no more than that, the array is not enough).

For 10 lakh calls Rand() overtakes the standard rand by 50 µs (time rnd.Rand()=344 time rand()=391).

Disadvantages: it takes a long time to initialize Srand(), standard range is 32768 (766 µs), 1 lakh almost 2 minutes.

I haven't checked the 10 million limit, but for sure it is incredibly long because time grows progressively.

Healthy criticism is accepted. I accept criticism in the currency of the deposit:)

Urain, stumbled on your post, this is an acute problem for me.

I'm not sure if the output will be really unique sequences with above 32768 sequences.

I won't claim that they will be similar exactly, but they will be based on those 32768 sequences.

That is, the output I will NOT get, for example, 2,000,000,000 sequences unique in their internal system.

I may be wrong.

I also did something similar, diluted random with random, the results with regard to the market were not very good.

I used dll, which gives billions of unique combinations, the results as applied to the market are more interesting.

But dll has one disadvantage indicator/expert can not be sold through the market and participate in the championship.

That's why I'm still in search of a mql randome with a large range.

I'd like to know what you think of my assumptions.

[Mykola Demko]

litechat:

Urain, on your post, this is an acute problem for me.

I'm not sure if the output will be really unique sequences with above 32768 sequences.

I won't claim that they will be similar exactly, but they will be based on those 32768 sequences.

That is, the output I will NOT get, for example, 2,000,000,000 sequences unique in their internal system.

I may be wrong.

I also did something similar, diluted random with random, the results with regard to the market were not very good.

I used dll, which gives billions of unique combinations, the results as applied to the market are more interesting.

But dll has one disadvantage indicator/expert can not be sold through the market and participate in the championship.

That's why I'm still in search of mql random with large range.

I would like to know what you think about my assumptions.

Is it a proprietary dll or an alien one? If it is, it may have a feature (like reading CPU temperature), and the rest may be replaced with MQL.

The chip is to ask the CD to be loaded into the nativ.

I don't see any other way.

[Rashit Ayupov]

Urain:

Is this dll your own or someone else's? If it is your own, then it probably has a feature (e.g. reading CPU temperature), and the rest can be switched to MQL.

The chip should be loaded into the Active.

I do not see any other way.

Dll is someone else's, there is source code for several pages.

I wanted to translate it all to mql, but realized that I can not.

No tricks because the sequences give the same and infinitely many, I have checked.

I just need random numbers that can be repeated more than once.

[Rashit Ayupov]

I think I found my dream library)https://www.mql5.com/ru/code/1146 write there is a good generator, I will try...

[revers45]

litechat:
I think I found my dream library)https://www.mql5.com/ru/code/1146 write there is a good generator, I will try...

And what's in the subject line can already be considered a pipe dream...?

[Rashit Ayupov]

And what is in the top of the topic can already be considered a pipe dream...?

Back when I didn't know what the Mwb exchange and forex were

I programmed neural networks, and not just any neural network, but a gold one.

There were a lot of networks and I gave them all the ability to cross and give descendants, all this worked by encrypting the structure of the networks into chromosomes

and the crossing triggered the usual mechanism of genetic mutation,

Basically I grew all sorts of creatures that adapted to the environment,

which consisted of good and bad zones with variable standing. In short they ran around the screen and tried to survive,

The input parameters were the receptors of the creatures and the output was the legs, their number varied depending on the result of the mutation (crossbreeding).

The results were good there were generations that adapted.

So that's what I'm saying, and when I found out about the stock market, about forex, I thought, this is the real application of neural networks,

and I spent three years programming networks for exchanges,

They did their job perfectly, they made forecasts.

But there was one point (butterfly effect) that I found out later, which was not profitable.

The slightest price change could dramatically change the forecast,

I realized that this instability of neural networks is not suitable for robots.

So I closed that topic for myself.

Anyway, this is just a story of my experience.

It all worked on a DBMS Cache + Delphi + API AlfaDirect