Market etiquette or good manners in a minefield - page 93

New comment
 
Neutron >> :

That's a fascinating fact. Ha! - It means that the Market does not like to move in one direction. After all, pattern 00 corresponds to two moves down and 11 moves up.

He doesn't like it, you bastard... >> Oh, he doesn't like...! Apparently it's because the market is in doubt (flat)

 

Ah, you motherfucker!

This is a picture of the states of the 3-input classifier. Can you see what the bastard likes?

It looks to me like the very visual representation of the market mood (state) that we are all engaged in visualising and interpreting here! This is the steady handwriting of the Market and I can finally see it!

 
The game seems to have turned out to be a negative MO
 

Although no.

000 - byu

111 - sell

001, 011, 100, 101 - smoking the bamboo

010 - sell

110 - byu

 
на 85 стр. M1kha1l писал(а) >>

It may be worth modifying the pattern from:

+1,-3,+2,-1

в

+1,-1,-1,-1,+1,+1,-1

But then several questions immediately arise, such as about the dimensionality of the pattern.

How do you decide if a pattern is long enough?

 
M1kha1l >> :

How do you decide whether a pattern is long enough?

The inputs are already binary, no +2 and -3 there. How to decide the optimum of the dimensionality of the NS input? This is also what you asked.... and Mathemat is missing somewhere, apparently fighting with the phoebes again.

 

The market does not like to move in one direction. 93 pages and such deep conclusions. Hmm, impressive:) Better gentlemen, tell me, how do you account for the diversity of views in the market? After all, not everyone in the market has zig-zag building systems. By the way, which also like to redraw:)

 
registred >> :

>> By the way, which also like to redraw:)


Ours don't. That's how we account for it.

 
paralocus >> :

Ours doesn't. That's how we account for it.

Everyone likes yours too, otherwise why would you need settings when teaching linearity of inputs for a non-linear model. The question, by the way, was not about technique, a rhetorical question, I'd say:)

 

paralocus wrote >>

The inputs are already binary, no +2 and -3 there. How do you decide the optimum of the NS input dimensionality? That's also what you asked.... and Mathemat is missing somewhere, apparently he's at war with the phoebes again.

Please note that:

  • the idea of binary inputs was proposed several pages before the discussion, as an alternative to the previously discussed patterns of segment inputs multiple of H
  • was suggested together with a question about length sufficiency, because with "short" patterns it's really not a problem to create their database, but from some N you need for example NS to interpolate "missing" patterns or a tool to bring a found pattern to one of known ( to a crystal node by Neutron)

I'll ask the question again in more detail:

  1. how to find optimal N-length pattern?
  2. predict on the basis of the found pattern:
  • stat.method (then it is necessary to have it exactly in the database) ? [what, imho, writes Neutron ]
  • or NS ?
  • or to "reduce" found to known and then stat.method?

Imho, obviously the answer to the first question depends, among other things, on the choice of the answer in the second.

Neutron: Maybe we should first reasonably choose a prediction method (see question 2)?

1...8687888990919293949596979899100...104
New comment