Machine learning in trading: theory, models, practice and algo-trading - page 1603

 
Evgeny Dyuka:
You can convince me that I am wrong. I worked on this project for 16 hours a day for six months and I got the result. I can not show you because you are waiting for a link that I will give will be advertising casinos.

A counter question, where else to discuss this topic as on the profile forum?
Maybe to be safe, let's equip a delegate from the forum, he will wear a protective suit, go to the link, observe and tell - is it a casino or not.
After it we will discuss.

What's the point of "discussing" if you don't want to show the data preparation process and the results of testing the system?

Discuss what?

 
Evgeny Dyuka:
The percentage of guessing is not important in this topic, there are other categories. What matters here is whether the network is trainable or not. If it shows signs of teachability then the door is open, further we can improve quality of prediction infinitely. Everything will depend on equipment and time.

Any network can be trained.

NSs are trainable. All

 
Dmitry:

What's the point of "discussing" if you don't want to show the data preparation process and the results of testing the system?

Discuss what?

I'm willing to tell you everything except what specific features are in the dataset for learning.
Here are the links, I hope I will not be banned for advertising:

1) Here I tell you a little bit about how I did it, link

2) This is where neuro signals every minute his prefix, link

3) Here I filter only good predicts for the public, link

 
Dmitry:

Any network can be trained.

NSs are trainable. All

I have not found a single working neural network, the topic is a complete failure
 
Evgeny Dyuka:
I have not found a single working neural network, the topic is a complete failure

Well, it depends on what is "workable"?

You wrote that the main thing is the teachability of the network - all NS are teachable.

 
Evgeny Dyuka:

I'm ready to tell you everything, except what exactly are the features in the dataset for learning.
Here are the links, I hope I will not be banned for advertising:

1) Here's a little telling how I did, link

2) This is where neuro signals every minute his prefix, link

3) Here I filter only good predicts for the public, link

There's no need for third-party links - you came to the forum, so tell me about it here.

Almost no one will follow the links.

 
Dimitri:

Well, it depends on what is "workable"?

You wrote that the main thing is the teachability of the network - all NSs are teachable.

Did you do this? It can't be the case that a neural network is trainable and doesn't show the same results in the real world
 
Dmitry:

There's no need for third-party links - you came to the forum, so tell me about it here.

Almost no one will follow the links.

Then let's take our word for it that everything works properly, that the network sees the trend and warns of a reversal, that predicates over 30 work fine.

Ready to answer technical questions

 
Evgeny Dyuka:
Have you ever done this? It can't be the case that a neural network is trained and does not show the same results in the real one.

Oh, so we are talking about the difference in network performance in training and in the test?

There's bound to be a loss - you can't do without it.

 
Evgeny Dyuka:

Then let's take our word for it, that everything works as it should, that the network sees the trend and warns of a reversal, that the predicates over 30 work perfectly.

I am ready to answer the technical questions.

Why take our word for it?

Show me the results of the tests.