NATURAL INTELLIGENCE as the basis of a trading system - page 97
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
Yes, judging by the rules, the conditions are close to real. But the question arises: what prevents the winner from opening a real account with $10K (like in the contest) and quickly doubling this amount using the winning Expert Advisor?
Well, we will never know, but we may assume that there is another account we do not know about. And the account that is publicly displayed on the viac just as a "general-information" )))) Why would you put an account you're really working on on public view? ))))
So talk about sensible systems
That's exactly what I'd like to do, as what's been so far is a conversation around rather than on the subject. I, for one, am interested in both the topic of 'intelligent systems' (whatever you mean by that) and philosophical reflection on life. Not to mention the possibility to communicate with normal people in their own language. :-) So, with your permission, in order.
The logic is unclear. What is the point of contrasting something to something, if neither this "something", nor this "something" has an objective definition. With this phrase you don't so much explain what you call intelligent systems as you further confuse. For that matter, you don't need any definitions. It is enough to show on concrete aspects, ways of functioning, properties of internal organisation of those and other systems, how and in what way they differ. If you can't do it because, for example, of confidentiality of information and non-disclosure of know-how, it remains only to illustrate by results or small touches like memory capacity. Unfortunately then there is generally nothing to discuss. Well, except to congratulate you on your achievements.
I believe that if a program can learn, with or without a teacher, to perform expedient and result-oriented actions, such a program can be called intelligent as well as a human.
Thus, NS that are able to learn both with and without a teacher, and even in the process of autonomous work at Forex, fall into the category of reasonable ones.
So what you've seen in the picture is only a confirmation of the theory in practice.
What do you mean - confirmation? In the sense that your program works? Well, there are a lot of working programs, including trading ones. In the sense that it may bring profit? You cannot see it on that picture. Not on this one, but on the one with the results. And again, I have profit-making systems as well, although they are not so numerous. And we are not talking about profitability, but about the reasonability of the system. And these things are quite different. Let me repeat: man is a reasonable system, but few people trade with profit.
My systems learn to trade, and as you can see, they do it quite successfully.
How does the learning process of your systems differ from that of NS?
In this case, I only construct a "genotype" and improve it by results of "natural selection". Just like in nature.
How is this different from genetic algorithms ? Is natural selection done by your hands or is it done by the system itself ? If we're going to talk, I'd like to see more substance.
You were given a link to look at it. I wonder if you knew the result beforehand. Does it make you smile?
If this is about him http://www.galactic.org.ua/Prostranstv/anoxin-1.htm. Could you tell us which of his ideas and principles you've managed to program?
Yes, it is about the ingenious Russian Neurophysiologist and his Functional System as interpreted by builders :)
However, I strongly recommend that you consult the primary sources.
In the attachment you'll find some of his work.
If this is about him http://www.galactic.org.ua/Prostranstv/anoxin-1.htm. Could you tell us which of his ideas and principles you have programmed?
To the second half of the question I can answer this way - all and a little more.
I would like to draw the attention of the audience to one very serious gap in most interpretations of the Functional System:
FS itself is just hardware for processing information, but the principles or software lie in the theory of Faster Reflection of the same genius author Peter Kuzmich and not less genius philosopher A.I. Oparin. And the power of the idea is limitless. With little help of imagination and using only remote understanding of dialectics of systems you may easily answer the question, what is "cunning" or "wise", how could a leader like Stalin or Lenin appear, how life was born and what is "intellect" (the latter will require at least superficial knowledge of physiology).
I have transformed the FS architecture a little, separating the Control and Execution subsystems with a slightly different architecture and a very substantial functional specialisation.
The principle of preserving FS architecture at all levels of organization is fully preserved.
In my programs two were used but only one level survived.
(I hope this paragraph has enlightened my "a little more")
And the rest - the principle of anticipatory reflection (in common parlance - the ability to predict the occurrence of biologically significant events on the basis of past experience), two subsystems based on associative memory, lots of receptors that analyse tick data, acceptor of results, etc.
Here it is simpler and less wordy - http://www.galactic.org.ua/Prostranstv/anoxin-2.htm
Excellent!
Note the links below: FS Theory - short but gives basic insights
Maybe someone would be interested - http://golovolomka.hobby.ru/books/gardner/mathdiv/14.shtml
The logic is unclear. What is the point of pitting something against something if neither "something" nor "something" has an objective definition. With this phrase you don't so much explain what you call intelligent systems as you further confuse. For that matter, you don't need any definitions.
It is enough to show on concrete aspects, ways of functioning, properties of internal organisation of those and other systems, how and in what way they differ. If you can't do it because, for example, of confidentiality of information and non-disclosure of know-how, it remains only to illustrate by results or small touches like memory capacity. Unfortunately then there is generally nothing to discuss. Well, except to congratulate you on your achievements.
Thus, in the category of intelligent get NS, who are quite able to learn with a teacher, and without him, and even in the process of independent work on the forex.
In what sense is confirmation ? In the sense that your program works? So there are many working programs, including trading ones. In the sense that it brings profit? You can't see it on that picture. Not on this one, but on the one with the results. And again, I have profit-making systems as well, although they are not so numerous. And we are not talking about profitability, but about the reasonability of the system. And these things are quite different. Let me repeat: a human is a reasonable system, but few people trade with profit.
So how is the learning process of your systems different from that of NS?
How is it different from genetic algorithms ? Do you do the natural selection by hand or does the system do it by itself ? If we're going to talk, I'd like more substance.
You were given a link to look at it. I wonder if you knew this result beforehand. Does it make you smile?
Dear Yurixx, I will try to answer in order:
The point of opposition is very significant, both scientifically and commercially:
The term "Artificial Intelligence" was discredited (discredited) with great success by previous impatient generations of cybernetics making scientific bets on Neural Networks, explaining on the fingers what it actually is (without strict definitions), and therefore it is almost abusive today and especially among traders, who back in the 70s tried to trade without the participation of the natural intelligence. There were successes of course, but mostly irreversible failures. Therefore I had to get creative when explaining to potential clients what my programs do :). Yes, that's why I was smiling - those who survived on market with Neural Networks are producing 7-12% yearly :)
Since I haven't heard about any major breakthroughs in this area in recent years, I took the maximum, and because my systems generate such profits in the first 15-20 minutes I felt like a "cool".
Forgive me - I relaxed.
On item 2: You assign me an impossible task because only P.K.Anokhin wrote the whole library and I have reread and digested it for 20 years. Precisely for this reason I strongly recommend to read it yourself. After that I will be happy to answer quite specific questions.
By giving my definition of reasonableness, which, by the way, does not claim to be scientifically rigorous or the truth in the last instance, I have explained to you what I am talking about.
If you find that NS falls under such definition, it would be silly to argue and probably the difference lies somewhere else.
If you read more attentively what I have written probably you would pay attention to sense of the offer - acknowledgement of my theory in practice.
As far as I remember I have never mentioned any results in this paragraph. And about a sensible man, losing on Forex - he, excuse me, does not fall under my definition :)
I'm sorry, I have to run, I'll continue at the next rush (the barbos is outside).
Maybe someone would be interested - http://golovolomka.hobby.ru/books/gardner/mathdiv/14.shtml
Very entertaining!
I agree that most likely the next step in the evolution of living things will be artificial "living things", obeying all the same objective laws of nature, and hence the highest goal of all living things - preservation of their own integrity (healthy natural egoism, self-preservation).
Imagine what will happen if a human being threatens the integrity of the new living? Because of the same desire and the law of "conservation of energy" (laziness) we deliberately create much stronger and more destructive machines than man, and you and I speculate on how to endow them with intelligence :)
How is the learning process of your systems different from that of NS?
How is it different from genetic algorithms? Do you do natural selection by hand or is it done by the system itself? If we are going to talk, I would like to have more substance.
How my method of learning differs from NS I honestly don't know. Because I do not know how to teach the NS (interested in it in the early days of my youth, but never gave much importance to the NS).
Despite the fact that your question again pulls on independent research work, at least at the level of theses, students of psychological or pedagogical high schools I will try to answer.
About the methodology I can inform the following:
I give systems to eat 200-500 days of data tick.
The general attitude is: if the market goes up - try to buy, down - try to sell.
If the result is positive when recognizing such a situation - do action andpriloschih a positive result.
If the result was negative - do the opposite action.
If that does not work = do nothing in this market condition.
Of course this list can be extended by stocks (Close, Limit, etc.)
I tried (tested) results of two approaches - the American and the Russian system of training (in the scientific: from private to general or from general to specific).
In other words: American way - learn very deeply, but not very widely, in translation to trader's language - use it for one day till there are no losses, then continue to the next day.
The Russian school: study wide, but do not dig deep (you will understand later). In other words, run through all available days and scroll through them until there are no losses.
The Russians win (as always).
In the first approach potential profits are erased in the learning process, because the set of possible winning signals is limited by the limited amount of information of one day.
But all other potentially profitable signals are not repeated after the first failure.
Modified initial conditions allowing a limited number of acceptable errors - the learning time increases exponentially.
The market behavior is characterized by very low activity and very unreliable profitability.
It is explained by the fact that it cannot recognize what it has not learned (the market is different in reality). Rejected.
In the Russian school there are no such restrictions and therefore works, but in the course of bidding to learn. (This is the conclusion for teachers).
One last thing: Yes, I kill them myself and being part of nature I make a "natural" selection for me.