Signs of a REAL system - page 2

 

Note: the last point in the 'fit'. This may also be an indication that the logic of the TS is correct in principle, but requires improvement - cutting off the "snail's tail".

Addendum: we can add to the same section the point about the dramatic difference in the results on different quotes histories (from different brokerage companies).

 
ForexTools писал(а) >>

it was right here.

Definitions from "neural networkers" are needed here. Neural networks have a phenomenal ability to remember. You always need to make sure that the thing has found a pattern, rather than having learned optimal input-outputs from history.

For example, a graph from a neuron's weights, if a pattern is found, looks more like a curve than a random set of points.

 
HideYourRichess >> :

Holy crap!


Yeah, well, systems that exhibit this behaviour are not experts. What's "awesome" about it? )))

 

ForexTools писал(а) >>

it was right here.

Ah. Now I know what you're talking about. I agree. This function approximation is one of the applications of neural networks. When applied to the use in EAs, it is a fitting.

 

I would add in 'Signs of fitting': difference in results at different intervals in history (for the same DC).

 

1.I wrote my addendum and comment on the topic above. There are more - I formulate them.

2. Without any irony, I think the topic is useful and deserves an article.

3. don't be so sacredly anxious about the topic you brought up - it won't get away with a few flood posts from adequate people.

 
joo писал(а) >>

Ah. Now I know what we're talking about. I see. This function approximation is one of the applications of neural networks. When applied to expert use it is a fitting.

Neural networks need to know how to prepare!

Now, on topic:

Fitting is a systematic failure to pass the Forward Test.

 
Mislaid >> :

The use of approximation in EAs is a fit. What does the skill of preparing neural networks have to do with it...(shrugging shoulders)

 
The ideal system in which only the lot size and the risk criterion are set in the external parameters...
 

By the way, niko1312 and Svinozavr have cool avatars)))

(see previous page)