Machine learning in trading: theory, models, practice and algo-trading - page 1503
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
Good time to all who are in the subject of NS.
I have not studied a lot of NS in the basic understanding, but there are questions and no one to consult with.
If anyone knows well the models of construction, please advise in what direction to look for a solution, and whether it makes sense.
The task is of the following nature.
There is a sample in a one-dimensional array, the values of the indicator being redrawn.
Is it possible to train the network using the NS to obtain the same array of samples, but without further redrawing of this sample in the real time?
That is, using a teacher to teach duplicate values of the sample, and will this result redraw in the real time?
If not, what model is better suited for this?
Clustering, as I understand it, gives the final answer as two solutions Yes No, true false, 0 1
That this learning model is not suitable for the task at hand.
What kind of learning model is needed for the task at hand?
And does it make sense to train to get rid of redrawing values in the final result?
Good time to all who are in the subject of NS.
I studied NS in the basic sense, but have questions and no one to consult.
If anyone knows the model, please advise me in what direction to look for a solution, and if it makes sense.
The task is of the following nature.
There is a sample in a one-dimensional array, the values of the indicator being redrawn.
Is it possible to train the network using the NS to obtain the same array of samples, but without further redrawing of this sample in the real time?
That is, using a teacher to teach duplicate values of the sample, and will this result redraw in the real time?
If so, what is the best model for this?
Clustering as I understand it gives the final answer as two solutions Yes No, true false, 0 1
That this learning model is not suitable for the task at hand.
What kind of learning model is needed for the task at hand?
And is there any point in training to get rid of redrawing values in the final result?
If the result is constantly redrawn, then what is the final result?
Why not count as everything else, by the close of the candle?
If the result is constantly redrawn, then what is the final result?
why not count as everything else, by the close of the candle?
We teach arange of values, not an extreme value.
The final result is the original teacher in the one-dimensional array.
I.e. the point is to copy for example the same mask, but that it is not redrawn in real time (the mask is for example, it does not draw)
I haven't studied much NS in basic understanding, but I have questions and no one to consult with.
...
Is it possible to train a network using NS to get the same sample array, but so that this sample does not redraw in the real time?
That is, with the help of a teacher to teach duplicate values of the sample, and will this result redraw in real time?
strange basic understanding of NS, and what then in your concept of NS training error?
The answer is no, the NS will always redraw according to your understanding, even if you trained on the data not redrawing, you can experiment with activation functions and the structure of the NS, but still "will" be stuck with the error of training the NS - this error will "redraw" when calculating the trained NS
like this
strange basic understanding of the NS, and what then in your concept of NS training error?
The answer is no, the NS will always redraw according to your definition, even if trained on the data not redrawing, you can experiment with activation functions and the structure of the NS, but still "will" be stuck with the error of training the NS - this error will "redraw" when calculating the trained NS
like this
Learning error is the difference between the desired and actual model output.
It does not allow you to evaluate the accuracy of the model with new data that were not involved in the training process.
It is better to use generalization error, i.e. model error on a test set.
So it is like this.
That's why I wanted to know if it is possible to solve this problem using NS.
And about possible models which can solve this problem.
Yes, the idea was just to take a range of non-drawing values from the past and train the network on this sample.
There are specialized programs for training with ready-made models or constructing them manually.
The question is which model is more suitable for the task, the name of the model, I am sure there is such a model.
Yes, perhaps there will be an error, but you can try to minimize it, the main thing that the model was chosen correctly.
I will listen to the opinions of other participants.
How's it going with the real data?
How's it going with the real data?
Seems to be pulling in the plus, but it's too early to tell. I just started it on my demo account on Friday. I am thinking to increase the minimum entry trigger to decrease the number of trades and improve their quality.
It seems to be in the plus, but it is too early to tell. I just started it on Friday on my demo account. I think the minimum entry trigger should be increased to reduce the number of trades and improve their quality.
It is a rubbish, not a TS, my friend.
It seems to be in the plus, but it is too early to tell. I just started it on Friday on my demo account. I think the minimum entry trigger should be increased to reduce the number of trades and improve their quality.
This is crap, not the TS, my friend.
With such a low number of trades you can't even recognize the Grail. The difference between the analytical results and the real ones is not so great, in this case the trader needs to re-calculate the probability of success.