Interpolation, approximation and the like (alglib package) - page 11
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
I just asked if anyone has done it to save time.
What a stupid question.
Why the aggression?
A question like that to me would give the answer "Who can do it?", because being able to work with someone else's code in another language and interpreting it into another language is not such an easy task, it seems to me.
Where is the aggression coming from?
A question like that to me would give the answer "Who can do that?", because being able to work with someone else's code in another language and interpreting it into another language is not such an easy task, it seems to me.
I just don't want to pour water on an abstract
I just don't want to pour water in the abstract
So in fact they've already poured in a decent amount of water, and so maybe there would be someone capable of implementing it.
The fact is that they have already poured in a decent amount, so maybe there would be someone capable of making it happen.
What's the point of this message? I have almost done it myself. It is not all interesting and there is essentially elementary
What is the point of this message? I've almost done it myself. Not everyone is interested in it and it is elementary in essence.
In the end, the proverb goes well with our dialogue "Do not do good, you will not get bad". I wanted to help you by giving a different wording to your needs, and ended up receiving negativity and resentment. Walked away.
In the end, the proverb "Don't do good, you won't get bad" fits well with our dialogue. I wanted to help you by giving a different wording to your needs and ended up with negativity and resentment. Walked away.
You wrote the same thing but rearranged the words and added some of your own experiences to the context ) As a result, the simple phrase "transformation of traits" went somewhere deep into subconscious images
I understood the goal of the project from the long time back. In fact, in the previous article of RDF I requested Mr. Maxim to IMPLEMENT, ,,, for for Something, Similar to change the function or strategy immediately upon market change.
But now Maxim is bringing too many new concepts to MQL like "kernel tricks" and "Splines" which are completely new to me though I have to learn to see if these have any advantages. Theoretically, both the methods seem to be very promising, but I am doubtful about the end results. If it does not, it's not enough to bring the consequences of reinforcement learning.
I have several simple ways of implementing. For example, simply using a basket of trading systems in the OnInit () function and using switch.
In my perspective for any machine learning algorithm to work in forex market, it must meet the following 3 criteria:
1.Speed of learning from the training data during optimization
2.Speed of calculation from the trained data at the time of trade entry and exit
Convergence of the algorithm automatically over time by repeated learning without showing random behavior
If the reinforcement learning algorithm fails to perform any of the above tasks, then it is just a simple algorithm for trial and error.
The previous article of Maxim about RDF was successfully able to complete the first 2 goals, but failed to complete the third task. the algorithm so that it can converge over time even though we can do the optimization manually once in a week.
But I am still hopeful that Maxim can complete using Kernel tricks and splines and I am eagerly waiting for the next article.
I have read the topic, and imho, the familiar style that has already dominated the forum - everyone speaks exclusively about his own, sore point, occasionally distracting to offend his neighbour.
And as to technique, if we speak about kernel tricks, mapping of feature space, it's clear to me only in respect to linear classifiers of reference vectors type, but why trees, forests, neural networks have their own kernels inside black boxes?
And if the task is just to make up external formulas, then symbolic regression and genetic programming will be cooler, in my opinion, than irrelevant interpolation and approximation).
Well, technically, if we are talking about kernel tricks, mapping feature space, it is clear to me, only in relation to linear classifiers like reference vectors, but why should trees, forests, neural networks have their own kernels, inside black boxes.
non-linear ones why offended?
but combinatorial kernels. The task is to reduce errors by small blood, which I have successfully managed to do. Build 150 layers in a neural network and then remove 1 layer/neuron at a time. It'll be worse and slower.
And I can train 10-20 thousand models in half an hour :)
non-linear for what offended?
but combined nuclei. The task is to reduce errors due to low blood, which I successfully obtained. Wind up 150 layers in the neural network and then delete 1 layer \ neuron. It will be worse and slower.
and I can train 10-20 thousand models in half an hour :)
Sounds great about the training speed :)
But are you sure that while trade entry the algorithm can check such large number of trained models so quickly to make an accurate trading decision?
I mean how much time gap do you expect on average between checking the models and trading decision in a regular VPS? Or are you going to implement a trading policy similar to your previous article?