Machine learning in trading: theory, models, practice and algo-trading - page 3619

New comment
 
Evgeni Gavrilovi #:

Already found it. Is the whole notebook with all the code or one parquet file sent for checking?

The whole notebook, but only functions at the very bottom are used during the check, other lines are automatically commented out and are not executed.
 

100 trillon to work out the weather astrologicals?


 
Interesting dataset. Got stuck last night. Tried new features based on convolution of graph edges, like in super-precision networks - no improvement. Tried some other things, also no significant improvement.

Either continue the topic with dip lerning rather than forest/busting, or there is another idea, secret for now :)
 
Maxim Dmitrievsky #:
Interesting dataset. Got stuck last night. Tried new features based on convolution of graph edges, like in super-precision networks - no improvement. Tried some other things, also no significant improvement.

Either continue the topic with dip lerning rather than forest/busting, or there is another idea, secret for now :)
Generate a trillion features and do fiche selection
 
mytarmailS #:
Geners trillion traits and do fiche selections
No such power and love the more thoughtful approach :)
 
Maxim Dmitrievsky #:
No such power and love the more thoughtful approach :)
How many signs are there in total?
 
mytarmailS #:
How many signs are there?
There are no signs. There are graphs with edges. The edges are described by numerical values, from which you should make attributes, or apply any other approach. All graphs have different numbers of nodes and edges.
So, you don't have to use learning with a teacher, but you can as an option.

The graphs in turn describe causal relationships between x, y and intermediate/separate nodes. There are 8 types of links in total.

You have to determine what affects what, what doesn't affect what, what interferes with what, and so on :)

They also write in their discord that you can change the graphs themselves. This makes the task even more complicated.
 
Maxim Dmitrievsky #:
There are no signs. There are graphs with edges. The edges are described by numerical values, from which we should make signs, or apply any other approach. All graphs have different numbers of nodes and edges.
That is, you don't have to use learning with a teacher, but you can as an option.

The graphs in turn describe causal relationships between x, y and intermediate/separate nodes. There are 8 types of links in total.

You need to determine what affects what, what doesn't affect what, what interferes with what, and so on :)

They also write in their discord that you can change the graphs themselves. This makes the task even more complicated.
So you need search and optimisation.

Take symbolic regression and write a programme to enumerate, rearrange the nodes and edges of the graph and do an evaluation.
 
mytarmailS #:
So you need search and optimisation.

Take the symbolic regression and write a programme of enumerations, permutations of nodes and edges of the graph and do the evaluation.
I'm not sure what's needed yet. When I start thinking about it, I realise I don't understand the point of the problem.

And Prado's kind of teasing: "you don't have to understand the task."

I'll have to talk to a good counsellor about it.

 
Maxim Dmitrievsky #:
I'm not sure what's needed yet. When I start thinking about it, I realise I don't get the point.

And Prado's kind of teasing: "you don't have to understand the task."

I'll have to talk to a good counsellor about it.

There's a chat room, talk to others like you.
1...361236133614361536163617361836193620362136223623362436253626...3741
New comment