Machine learning in trading: theory, models, practice and algo-trading - page 1930
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
There are three variables A, B, C. Some condition has been manually made from them.
I want to reproduce this condition automatically. There is no need to find it, because I already know it. But I need to have, for example, a dozen of some weighting coefficients, a certain combination of which can fall into this condition with high probability.
If the condition is known, then a dumb GA can find values A, B and C, at which it is satisfied.
And if not, then the problem has no unambiguous solution, you can come up with many different conditions for different values of the variables.
option 3 - through logistic regression + polynomial function generator. Then the output will be quite short and easily interpreted formula like yours
But I don't sprechen what is the point of reproducing the condition through MO
option 3 - through logistic regression + polynomial function generator. Then the output will be quite short and easily interpreted formula like yours
But I do not sprechew the meaning of reproducing the condition through MO
then it's "not fershtay")))
Well, neither do I.
A question from a beginner.
There are three variables A, B, C. Some condition is handwritten from them. For example.
I want to reproduce this condition automatically. I don't need to find it, because I already know it. But I need to have for example a dozen of some weight coefficients, a certain combination of which is able to hit this condition with high probability, when I set A, B, C there (polynomial or HC - I don't know, because it's zero) and get the original condition.
I am interested in what kind and how many input weights the desired function has, so that such original conditions can be reproduced via weights?
The grid with 1 hidden neuron is capable of learning the intersection of ma100 with price.
These are its parameters, 103 in total, the increments of the last 100 bars are fed to the input. Neuron with tanh activation is function tanh(x*W+b), basically in brackets linear regression, x is neuron input, W is matrix of weights, b is offset. In the hidden neuron: W of 100 elements, b of 1. And in the output: W is 1 element, b is 1 element.
"Another test example, the intersection of ma and price."
https://www.mql5.com/ru/forum/86386/page1879#comment_17365998
So, how to train trees by clusters, I tell you and show you.
clustered traits will always work worse than non-clustered ones! by clustering you inevitably delete some information.
There are a hundred examples, we combined them into a cluster - where is more information? in a hundred examples or in one number (cluster number)?
How do you take your dataset with predictors and closing price and load by specifying the column with the closing price, rather than using the option of generating indicators in R ?
As I understand it, since the target is ZZ tops, then the predictor part of the sample should be filtered out, here, and so to feed the predictors you have to filter out the predictor table the same way, or what?
Well you know how to load data, I gave you an example where I read your csv file, what exactly is the problem?
Exactly the same target - it has a sign ZZ not reversals, a typical target that we used to rape
A question from a beginner.
There are three variables A, B, C. Some condition is handwritten from them. For example.
I want to reproduce this condition automatically. There is no need to find it, because I already know it. But I need to have for example a dozen of some weight coefficients, a certain combination of which is able to hit this condition with high probability, when I set A, B, C there (polynomial or HC - I don't know, because it's zero) and get the original condition.
I'm interested in what kind and how many input weights the desired function has, so that such original conditions can be reproduced via weights?
Also a nerd, but interested in the problem).
Let us imagine A, B, C as randomly modified instances of an abstract object in some single-type environment. Observing A, B, C, we distinguish their common-significant (peculiar) property, the values of which change for all three on environment events, but differently (depending on different states of the environment). We feel the regularities, but cannot formulate them. However, let us say, at the event of a new instance, there are sharp bursts of values A, B, C, with preservation of certain proportions, and the environment expands. We like this pattern and want to describe the condition, but the environment is full of other objects and properties whose behavior is also mysterious and unpredictable. Then we decide to create a system that marks various regularities of parameters of various objects on the set of environment events. The system must record the minimum number of maximum meaningful links of objects (type of objects, properties, values + occurred event) and record them in the array, so that further under these conditions we could build algorithms of behavior artificially introduced into the environment synthetic-fabricated parametric complexes.
Can networks do this?
clustered traits will always work worse than non-clustered ones! by clustering you inevitably delete some information.
There are a hundred examples, we combined them into a cluster - where is more information, in a hundred examples or in one number (cluster number)?
I cannot agree with you, the information was not deleted, but on the contrary - systematized, in fact, instead of the original 3 classes we got 3*4=12 new classes.
Of course, the sample was conditionally reduced by 4 times, which made training more difficult, but with large samples this effect should be less significant.
I think we can try not to remove the whole sample, but to fill it with zeros in the part related to other clusters.
Another option is deeper learning, but I think the result will not be very good because of the small sample.
Intermediate result of clustering - increase Recall (completeness) by 34% and a drop in precision (accuracy) by 5%, ie greater generalization of data occurred, which may be useful in some cases.
Exactly that target - there is a ZZ sign not reversals, a typical target that we used to rape.
And I thought there were U-turns!
However, you still haven't answered the question earlier - is it possible to automate the saving of the picture of the built space? I want to cycle through, say, the same parameters ZZ and see how the graphical models change, in general to automate the enumeration of targets. I don't have time to sit at the computer.
However, you still did not answer the question earlier - is it possible to automate the saving of the picture of the built space? I want to cycle through, say, the same parameters ZZ and see how the graphical models change, in general to automate the enumeration of target. I don't have time to sit at the computer.
I do not know if you do it in 3D, I doubt it, if you do it in 2D, you can sort it out in R-Studio, but why do you need this nonsense!!! It's much clearer, simpler, better just watch the error in the new data, the output is not an abstract picture, but a specific number that you can work with, compare, sort, and so on
I do not know, if in 3D, then unlikely, if in 2D, like you can save from the R-studio, but why do you need this nonsense!!! much clearer, easier, better just watch the error in the new data, the output you have not an abstract picture, a specific number to which you can work, compare, sort, etc.
What number - have not noticed before - what is the meaning of the indicator?
A picture always gives more information than just a number.
I'm trying to run the code, which was sent yesterday.
I do not understand what is the syntax
What is the number - didn't notice earlier - what is the meaning of the indicator there?
A picture always gives more information than just a number.
I'm trying to run the code you posted yesterday.
I don't understand the syntax
because my data is called "d" ))
Type in your
clos <- dt$your close price