Machine learning in trading: theory, models, practice and algo-trading - page 565

 
Yuriy Asaulenko:

The modern concept of OOP implies that you can (or even shouldn't) know absolutely nothing about the inner workings of objects. Only the interface.

So not knowing the inner workings of an electric kettle does not prevent its use at all.

Usually such software is well documented, already tested by thousands of users and there is no doubt about its performance.


I disagree. Many times I improved some methods, added something of my own, modified it, etc. How could I do it if I didn't know the algorithm?

 
forexman77:

Thank you. It will be something to study.


By the way, their RF algorithm is analogous to this https://www.salford-systems.com/products/randomforests#random-forests

It is more for me, I will be interested to read it tomorrow.

If you really understand and good at programming - read more about the RF feature importance, the current version doesn't have it... I'd really like to add it :)

Random Forests®
Random Forests®
  • Salford Systems
  • www.salford-systems.com
We suggest the following minimum and recommended, system requirements: 80486 processor or higher. 512MB of random-access memory (RAM). This value depends on the "size" you have purchased (64MB, 128MB, 256MB, 512MB, 1GIG). While all versions may run with a minimum of 32MB of RAM, we CANNOT GUARANTEE they will. We highly recommend that you follow...
 
Forexman77:

By the way, even without a neural network my curvature on the futures is not much worse than yours.


In general, the main thing is the result, not the process.) The main thing is what you got, and how you got it is another question.)
 
Yuriy Asaulenko:
In general, the main thing is the result, not the process.) The main thing is what we did, and how we did it is another question).

Opera theater... :D

 
Yuriy Asaulenko:
Generally, the main thing is the result, not the process). The main thing is what we got, and how we got it is the second question).

Well, it's not all so rosy there are places where the advisor clearly begins to pour or trample and these intervals of 1.5-2 years.

And this is 300*24 is already 7200 rubles only for maintenance of the account for two years. Or how much is the subscription fee for the account below 50 thousand rubles.

 
forexman77:

Disagree. Repeatedly improved any methods, added something of my own, modified it, etc. How would I do it without knowing the algorithm itself?

Well, that's an amateur thing to do).

As Knuth wrote in "The Art of Programming" - A good programmer must be able to optimize and improve any program, and Never do it.

 
I'll just leave it here.
Artificial Intelligence, AI in 2018 and beyond – Towards Data Science
Artificial Intelligence, AI in 2018 and beyond – Towards Data Science
  • 2018.01.03
  • Eugenio Culurciello
  • towardsdatascience.com
Or how machine learning is evolving into AI
 
Yuriy Asaulenko:

Well, okay.

There's a story. We took the NS (MLP), and we want it to find entries into deals. The algorithm of transaction support is known. Where when and how to enter trades - complete uncertainty. The task is to make a profit.

Your actions. With what and how will you teach? What, by the way, are you going to enter?

Read here, here, here, here, here, here, here. Maybe your questions will go away.

What kind of history are we talking about? Quotes, deals, or ...?

2. "The algorithm of transaction support is known" - what kind of algorithm is this? A simple example.

It's impossible to discuss empty words. Draw at least a simple structural scheme, so that we can understand what you are talking about.

Good luck

 
Vladimir Perervenko:

Read here, here, here, here, here, here, here, here. Maybe your questions will go away.

What kind of history are we talking about? Quotes, deals, or ...?

2. "The algorithm of transaction support is known" - what kind of algorithm is this? A simple example.

It's impossible to discuss empty words. Draw at least a simple structural scheme, so that we can understand what you are talking about.

Good luck

I see. You do not understand the tasks in general terms.) I do not need to read anything, the NS has already been working for a couple of months. I am interested in your approaches.

Let's repeat the condition of the problem. Take any input data you need - publicly available history, maintenance algorithm, and so on. Without limitation).

How will you solve the problem? What will you do? You don't have to actually do or solve anything. Just a description.

 
Yuriy Asaulenko:

I see. Tasks in general form you do not understand.)) I don't need to read anything, the NS has been functioning for a couple of months now. I am interested in your approaches.


You follow the links and read all my approaches. With examples, codes and explanations.

The dispute is over.