Using artificial intelligence at MTS - page 14
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
Interesting point. It is enough for a man to tapping on the keyboard a sequence of poorly connected words, abundantly and inappropriately sprinkle them with commas, dashes and other special characters, colour all this in different colours, and it already means that he is a philosopher? A peculiar idea of philosophy ...
These words are known to everyone and are used by many. And the fact that they are, among other things, philosophical categories worries very few people. However, IMHO, one should not see philosophy everywhere where these words occur.
And I personally understand the abstract level as the level of maximum generalizations. I did not see generalizations in usdeur, but I saw abstractness in another sense - in the sense of detachment from any concrete things. Particularly the ones under discussion on this forum.
Finally, in response to my post, dear usdeur finally explained what he wanted.
My problem is that I know how, but do not know with whom to have the opportunity to INTRODUCTORY IDEA.
Oh well, anyone who wants to give it a try. Especially since, as he wrote about himself
I am a trader KNOWING how to distance myself from SUBJECTIVES, from unnecessary inaccuracies.
Others who need it can find the address.
I won't look for any more of your addresses, comrade coder. But I would like to apologise for the label of language illiteracy.
If I were a moderator I'd delete all that colourful "nothin'" or maybe nothin', as it clutters up an interesting topic. But that's just a word....:)
And if matter of fact, I have been trying to give a bit of reason to EAs for a long time and with varying success. I have divided them into two groups:
1 - Advisor that implements a strictly laid down algorithm in accordance with the parameters given to it ( All of the basic parameters are usually found using the optimizer, such as Stop, Take, boundary values of indicators, parameters of indicators, etc.)
2 - Advisors that realize their algorithm in accordance with parameters changing depending on the market ( Optimizer is needed here too, but it is chosen certain ratios, depending on the basic parameters of execution and market characteristics, such as percentage of daily channel for Stop, Take, etc. Reshetov's Expert Advisor pushed me (frankly speaking, I do not understand what pushed it, because there is nothing similar) to create another type of Expert Advisor:
3 - the Expert Advisor that does not have a rigidly specified algorithm (for example, the usual conditions of opening a deal like "If A>B=>Buy" are absent in this EA, what are A and B, < or > and Buy or Sell, the EA decides for itself (it has been implemented only this way so far)), i.e. the EA is equipped with some kind of a Sell or Buy condition. i.e., the Expert Advisor equips itself with some set of tools, interprets the results of these tools / simultaneously selects the desired tools from the set of tools, and the Expert Advisor tries to act.
The task, of course, is not easy, given that the "thoughtfulness" (1 run in the tester takes about 1 hour so far) of the EA rejects the possibility of using the optimizer, the use of which would be useful at the stage of creation of the EA, in the future when using the EA, the optimizer should not be needed. I had to solve a few more tasks along the way and bring in some more experts... But the results are very encouraging. Unfortunately, so far good results are seen with low activity of the Expert Advisor, and attempts to "force" it to trade more actively have led to somewhat weaker results, but I think this can be fixed. If anyone is interested I can post some reports in the evening, I am not available at the moment.
- A couple of times in my life I have come across such sectarian callers and both times the situation was quite comical - I remember with a smile. The first time was a long time ago - I was on the bus, deep in thought, and then from somewhere on the side there appeared
a propagandist for the white fraternity of black Satanists or something like that. At first I was politely listening to his jabbering, but then he started to bore me and I blurted out something like - yes I would come to your meeting, but my worldview-axiomatic set doesn't allow having creationist constructions. The guy went silent, his eyes
opened wide and you could read in them, "So crazy! Do we need a crazy person in the sect?" Then he seemed to decide
- he might be good for something, he hesitantly suggested to me: "Well, you... come anyway.
The second time, it was a Sunday, there was a knock at the door. My wife opened the door - she's an extremely intelligent, but irritable
and impatient. She particularly hates it when the person she is talking to starts pounding water in a barrel. But here she was in a good mood this morning - she was frying some pancakes, and here was this agitator. He is a respectable fellow, as soon as the door is opened,
in a leisurely, measured bass, began to speak a memorised monologue about a bright future. She was listening to it good-naturedly for about three minutes, then interrupted him halfway through with a good-natured exclamation - "Shorter!" The man was not expecting such a reaction and his programme was going off and he started repeating the monologue from the beginning, but at a faster pace and in a higher voice.
Having listened to the same place, his wife interrupts irritably - "In brief!" The man again begins to repeat the monologue word for word from the beginning, but already in a falsetto. When she has reached the same point she boils over and shouts "In brief". She slams the door after the man starts babbling on at an incredible speed, starting all over again. I'm in the next room listening to the whole thing and laughing - hiccups ensue.
2 - Advisors implementing their algorithm in accordance with parameters changing depending on the market (Here, too, you can not do without the optimizer, but there are selected certain ratios depending on the basic parameters of the execution of the market characteristics, such as percentage of daily channel for Stop, Take, etc.)
For six months I have been studying these two groups of advisors, personally, I do not doubt the greater potential of the second group.
In some local thread I saw reports that over-optimization can even be put in the body of an EA.
3 - EA without a rigidly written algorithm (for example, the EA does not have our customary conditions for opening a deal of the type "if A>B=>Buy", which is A and B, < or >, and Buy or Sell, the EA decides itself (so far this has only been implemented)), ie the EA is equipped with some set of tools, interpreting the results of these tools / simultaneously selecting the desired tools from the set, the EA tries to act.
- A couple of times in my life I have encountered such sectarian callers and both times the situation was quite comical - I remember with a smile. The first time was long ago - I was on the bus, deep in thought, and then from somewhere on the side there appeared
a propagandist for the white fraternity of black Satanists or something like that. At first I was politely listening to his jabbering, but then he started to bore me and I blurted out something like - yes I would come to your meeting, but my worldview-axiomatic set doesn't allow having creationist constructions. The guy went silent, his eyes
opened wide and you could read in them, "So crazy! Do we need a crazy person in this sect?" Then he seemed to decide
- he might be good for something, he hesitantly suggested to me: "Well, you'll... come anyway.
The second time, it was a Sunday, there was a knock at the door. My wife opened the door. She's extremely intelligent, but irritable
and impatient. She particularly hates it when the person she is talking to starts to stir the pot. But here she was in a good mood this morning - she was frying some pancakes, and here was this agitator. He is a respectable fellow, as soon as the door was opened,
in a leisurely, measured bass, began to speak a memorised monologue about a bright future. She was listening to it good-naturedly for about three minutes, then interrupted him halfway through with a good-natured exclamation - "Shorter!" The man was not expecting such a reaction and his programme was going off and he started repeating the monologue from the beginning, but at a faster pace and in a higher voice.
Having listened to the same place, his wife interrupts irritably - "In brief!" The man again begins to repeat the monologue word for word from the beginning, but already in a falsetto. When she has reached the same point she boils over and shouts "In brief". She slams the door after the man starts babbling on at an incredible speed, starting all over again. I'm in the next room listening to the whole thing and laughing - the hiccups begin.
Yurixx 12/14/2006 1:13 PM -- ANSWERED clearly!
And now, after that, to WRITE such a message?
It shows the power of desire - to dilute!
But what to dilute? And most importantly with what?
And more importantly understand - WHY?
I came here to give, and to get help myself.
If I don't give a damn about the CRIkunov program, it should not be taken lightly .
IKING, on the other hand, should be taken elsewhere.
But a reasoned objection to what I WOULD have written would have done more good ! For me and for ALL FORUMANS.
When you don't do it, you get threats and hiccups like this... :(
The question to Reshetov, as one of the advanced "neuro-networkers" who came to communicate with traders!!!!. So to say, while my computer is down, I am optimizing my new grid :) A question - what do you think about something similar (from here http://forum.basegroup.ru/message.php?id=6620).
Neural networks? Forget about them, it's not even yesterday, it's the day before yesterday when computers were very slow. Classic neural networks are now a relic of history. An antique, not a tool. A monument.
Chewing over neural networks in runet is mockery and misinformation, like teaching calculus in computer science class.
Killed a lot of time on it myself. It's all been dead for a long time. A very long time ago.
Go straight to
http://www.kernel-machines.org/
http://www.gaussianprocess.org/
http://www.boosting.org/
See NIPS conference archives (books.nips.cc), ICML, sit JMLR (jmlr.csail. mit.edu).
And dig at citeseer.ist.psu.edu (not the basegroup)
The efficiency of classical neural networks? The whole basegroup will spend a month tuning parameters, it will spend hours training those networks, and still it won't get the results that the state of the art algorithms get in minutes and completely automatically, without any user-defined values.
Large-scale is the subject of today's research. There are no problems with small samples.
I'm not just talking about the website, I'm talking about information on runet in general. This level is a mockery of people. You have to try so hard to ignore normal methods and stubbornly describe all sorts of surrealism. More than 10 years ago we wrote tutorials for retired nerds on calculation of linear models, LOOCV considered analytically and computationally cheap, on regularization and forward selection, on orthogonalization. Where is it in runet, where? (though maybe it already exists somewhere) After all, all this is simple, and will beat BP with all its moments both in quality and speed. I mean, Chen et al are still on the cutting edge of basic pursuit, and get great results. And that's without any complicated theories, just know how to rotate matrices.
Although, if I understand the "letters" correctly
Suppose we have MLP and non-MLP, and ways of training A and B. Suppose in tests MLP and non-MLP trained by method A give consistently about the same high results, and MLP and non-MLP trained by method B give consistently about the same low results. And what should be the definition of the method then? MLP/non-MLP or A/B ? Roughly, this is the present situation. The difference lies precisely in the principles of the "training" algorithms, which do not care whether they are tuning MLP parameters, RBF network, polynomial or just a linear combination of inputs.
I can not interrupt optimization either in NSDT or in TS for the moment :)
It is clear that from simple to complex, everything should be done in your own way, but all the same, you can not learn and study everything, and you want to earn money and other bourgeois benefits. :)
Not so much as not to get a crust of bread, and the 3rd deposit is still holding out :)
...A question to Reshetov, as one of the advanced "neural networkers", who has condescended to communicate!!!! with traders. So to say, while the computer is down, I'm optimizing another grid :) a question - What do you think of something like this (From here http://forum.basegroup.ru/message.php?id=6620)
I understand that it goes from simple to complicated, you need to understand everything with your own brains, but you still can't know and study everything, and you want to earn money and other bourgeois things. :)
Not so much as not to get a crust of bread, and the 3rd deposit is still holding out :)
Test on a representative sample and outside the sample, and then look at the balance figure, it is much more accurate will tell you more accurate information closer to reality.
3 - EA without a rigidly written algorithm (for example, the EA does not have our customary conditions for opening a deal of the type "if A>B=>Buy", which is A and B, < or >, and Buy or Sell, the EA decides itself (so far this has only been implemented)), ie the EA is equipped with some set of tools, interpreting the results of these tools / simultaneously selecting the desired tools from the set, the EA tries to act.
No one is arguing that it's difficult, but not impossible. Of course, it may not be the best, and the algorithm is very simplistic, and written poorly, and works slowly (they're working on it now). - but it works. Results are not outstanding, but considering that actually forward test and new approach for me, I think it's quite decent: