NATURAL INTELLIGENCE as the basis of a trading system - page 101

New comment
 

1. Рецпторы - преобразующие разнокачественные воздействия внешнего и внутренних миров системы в последовательности каких-либо сигналов. Слух,зрение,обоняние и тд. преобразуются в эелектрическую импульсацию,а например в бизнес системах это отдел маркетинга,R&D и тп., производящие всевозможные отчеты...

If it's not a secret, please explain this point "on your fingers"! If a researcher can identify combinations of indicators or prices that predict the next market movement with >50% confidence, then it's not that hard to power all this magic to a control system (neuro, fuzzy, neuro-fuzzy) and get profits. Where to look for these combinations?

There is one more variant, when a forecasting system makes prediction of the next price movement and then sends it to controller, which makes control according to the forecast. In this case you have to work with prices, i.e. transform them, because price time series are very poorly predicted... It seems to me that we need to move away from the time dimension somewhere... But where? :)

From my little experience I can say that main "know-how" lie in transformation of prices or their indicators, and application of artificial intelligence allows to make more flexible decision-making or prediction system, but in most cases you can (I think so) check idea for prognosticability with usual recursive linear filter and if it doesn't give results, you shouldn't waste your time with more abruptness. I'd be very happy if anyone can find significant errors in my views and provide evidence to that effect.

 
renegate писал (а) >>

If it's not a secret, please explain this point "on your fingers"! If a researcher can identify combinations of indicators or prices that predict the next market movement with >50% confidence, then it's not that hard to power all this magic to a control system (neuro, fuzzy, neuro-fuzzy) and get profits. Where to look for these combinations?

There is one more variant, when a forecasting system makes prediction of the next price movement and then sends it to controller, which makes control according to the forecast. In this case you have to work with prices, i.e. transform them, because price time series are very poorly predicted... It seems to me that we need to move away from the time dimension somewhere... But where? :)

From my little experience I can say that main "know-how" lie in transformation of prices or their indicators, and application of artificial intelligence allows making more flexible decision-making or prediction system, but in most cases you can (I think so) check idea for prognosticability with usual recursive linear filter and if it doesn't work, you shouldn't waste your time on more cool stuff. I'd be very happy if anyone can find significant errors in my views and provide evidence to that effect.

Indicators are only part of circumstantial afferentation. Moreover, it is always the delayed part. That's why it has the meaning of the context in which market events develop (it's important for making decisions). Another part of situational afferentation involves assessment of the current state and forecasting of the nearest future. In addition, changes in internal state may serve as elements of circumstance afferentation.

In addition, in real life, the NEWS component is very important and can completely change the course of predicted events and all plans.

Where to look for them is in the price itself. Look for ways to describe all three components :)

You are absolutely right on the second and third point except for the last part of the third.

Time series and price curve patterns are very poorly formalised and predictable especially in real time.So mathematical methods are not effective. Therefore they are not used in my systems.

Basically the question of the receptor function can be formulated as follows:

How to encode the price curve invariant to the price itself for the subsequent extraction of signal features (trigger stimuli) and their recognition in real time?

How to use the information contained in the sequence of these codes to predict and evaluate the current situation ?

The answer to the first question depends on your implementation.

Read up on Reality Reversal and you'll be able to answer the second question.

Trend = those predicted but not yet realised prices - a more or less long term forecast.

The first arrow - Trend - the established direction of price movement.

The second arrow is a signal of a possible trend change.

The third one is an indicator of the current price position in relation to the limits of the nearest price range

Yes, as long as we are here:

No Sell = 100% certainty of the system that the price keeps going up

May Buy = You have the right to buy, our system was buying and winning - try it :)

Prediction History = expected price and trend line extremes at the official market close

(probably not a good term)

>> Good luck!

 
Yurixx писал (а) >>

Ah yes, sorry, that makes a fundamental difference. As opposed to the SWOE, by playing SMEs you will of course make everyone rich. No doubt about it.

Of course, if you have nothing to say, not even Comrade Integer can help.

When I asked you questions, which may arise in not too experienced person while getting acquainted with your system, you couldn't find anything better than to copy extracts from your talented and super brief help. It was a waste of effort, direct answers would have been much shorter and more informative. Or maybe you just don't know the straight answers?

Maybe that's why afterwards, when I still kept asking, you framed your nephew instead of answers ?

The only question you dared to answer was why you are selling a "money printing machine" instead of just printing it. You chose it yourself "for a separate answer", which I suppose illustrates its importance. Unfortunately that response, coupled with your response to a couple of my very humble comments, contains some obvious absurdities. And where does my attempt to move from your general phrases to the specifics of your own position lead to ? You're framing the other again, this time by Integer.

Well, I did not speak about help, but about quite different, though not philosophical, but quite systemic issues?

All right, let's stop here. I won't pester you with my questions anymore.

In conclusion of this interesting conversation one last one.

Does your system allow full automation or allows using it in hand trading?

I think this quote from the help file answers your question:

-------------------------- цитата------------------------------------------------------

For experienced traders:

1. We consider the "Trend" tool to be the most important one for confirming your own strategy, expectations or signals.

2. thePredictions table is used to quickly re-evaluate the outcome of your expectations if the market goes against you, enabling you to make a more efficient entry/exit.

3. For automated trading systems, we provide an API which can be used to generate signals, as well as to evaluate your own system-generated signals.

---------------------- конец ---------------------------------------------------------------------------------

I think its interpretation is up to you.

But to scholasticism like ". from the point of view of banal erudition ..." I do not answer on principle, especially as there is no question..., there is a completely meaningless set of, excuse me, stupid and contradictory judgments...

Ask substantively, after a VERY INTENTIVE ANALYSIS (highlighted specifically for you) of my statements, read the recommended literature (links), then it will be interesting to discuss with you (there will be an object of discussion), trying to find out the truth.

Don't be offended, take note and ask.

Yes, just found it:

'All ingenious is simple or another Grail?'

Last comment...

 
Here. To confirm my thoughts. I recently discovered for myself (quote from here):

"...One can learn to feel the market and predict it without any technical aids. Everyone can do it with proper training. To do this you need at first just observe the price movement on a minute chart of any chosen instrument (but only on one and the same, because each has its own specifics, and until you learn to feel it - work with one). It is difficult to say exactly how long it will take, it is strictly individual, for some it will take half a year, for others maybe a year [...] Important detail, you should not try to guess the price movement at this time, because trying to guess you distract your subconscious mind from work, and everything happens at the subconscious level.

The mind tries to guess, but it cannot look into the future, it only hinders the process. It should be very calm, focusing its attention on price movement and not thinking about anything else. Just observe, trusting yourself and having patience. The price charts are just symbols, through which contradictory and multipolar events in the world economy are expressed. Price charts as symbols help you tune in to those energies which shape and direct the movement of the market.

These energies appear and their direction is shaped long before the price changes on the chart itself. By focusing your attention on the chart you will subconsciously connect to these energies, you will start to feel them. And gradually a sense of knowledge will come, it's exactly like knowing where the next tick will go a moment before it happens...".
 

There is a publicly available plethora of damaging techniques posted by good people specifically for Pinocchio seekers.
However, let's read the epigraph carefully.
(I underlined something there))))

..........

"The flawed becomes perfect, the crooked becomes straight,

the empty becomes full, the old is replaced by the new;

"the striving for the little is the attainment of the much;

the pursuit of much leads to delusion.".

Lao Tzu "Tao De Jing"

....

Well, in general: - Trader! Watch out for personal psycho-physical safety!
...in today's environment, the border is the threshold of your home (c).

 

P.S. Pressed to the right edge: the forum editor is glitchy - unable to change the formatting of the original document.

 

Yes Implex, you are referring to an article without a single formula or line of code. Just for natural intelligence, so there's no maths there at all :)

 
Mathemat >> :

Yes Implex, you are referring to an article without a single formula or line of code. Just for natural intelligence, so no maths there at all :)

I've dug through a lot of information on neurocomputer modelling in my time (which I think has a bit to do with mathematics), studied and even implemented, albeit in a limited MQL environment, a multi-layer neural network. I got so fed up with formulas and "lines of code" that at some point I completely forgot what it was all about... I wrote a script-neural network, which allows you to conduct training and simultaneously observe the results of training on a test sample. And it's directly in the process of training. Over 600 "lines of code".

If you're interested - this is what I was doing before:

//+-----------------------------------------------------------------------------------------------------------------------+
//|                                                                                                          NeuroNet.mq4 |
//|                                                                                                             ImplexLab |
//+-----------------------------------------------------------------------------------------------------------------------+
#property copyright "ImplexLab"
#property show_inputs
//+-----------------------------------------------------------------------------------------------------------------------+
// блок переменных на глобальном уровне
extern bool teaching_true__use_false = false;
extern bool continue_training = false;
extern bool multiple = false;

int
 layers_count = 5,                        // количество слоев
 layers_neuro_count[] =                   // количество нейронов в слоях
  {20,2,3,4,3},                           // {кол-во нейронов в первом слое, кол-во нейронов во втором слое и т.д.}
 inputs_count = 50,                       // количество входов первого слоя
 outputs_count = 1,                       // количество выходов, должно быть равно кол-ву нейронов в выходном слое поделенному на два
 quantity_of_example = 100,               // количество примеров обучающей выборки, диапазон инициализации значений
 quantity_of_repetitions = 1000000,       // количество повторений (практически должно быть равно ~)
 inputs_index = 1000,                     // индекс начала входов (заодно и рисования результатов) для использования сети
 prediction_depth = 20,                   // глубина известно чего
 frequency = 10,                          // частота обновления визуализации ошибки
 repeat_count = 200,                      // количество повторений для функции analysis()
 shift_concerning_a_beginning = 1000,     // сдвиг относительно начала
 cntr=0,                                  // счетчик эпох
 num_input,                               // вход_из_обучающего_множества
 some_variable;                           // какая-то переменная (неизвестно зачем)

double
 xt[210][10010],                          // x[конкретный_вход][вход_из_обучающего_множества] t - traning, u - use
 yt[210][10010],                          // y[конкретный_выход][выход_из_обучающего_множества]
 xu[210],                                 // x[конкретный_вход]
 yu[210],                                 // y[конкретный_выход]
 w[10][210][210],                         // w[слой][нейрон_слоя][конкретный_вес]
 outt[10][210],                           // out[слой][нейрон]
 outu[10][210],                           // out[слой][нейрон]
 s_err[10][210],                          // s_err[слой][нейрон]
 net = 0,                                 // переменная net для многократного использования
 max,                                     // контейнер для максимальных значений входов
 min,                                     // контейнер для минимальных значений входов
 speed_of_training = 0.1,                 // коэффициент шага изменения весов
 weights_init_range = 2;                  // диапазон случ. нач. величин для весов, -weights_init_range<w<weights_init_range
                                          // при weights_init_range=2, нач. вел. будут лежать в диап. -2<w<2

//+-----------------------------------------------------------------------------------------------------------------------+

//+-----------------------------------------------------------------------------------------------------------------------+
// Список функций (17)
// 
// double neuronet_training()
// double activation_func(double net)
// double value_get_for_training(int index)
// double value_get_for_use(int index)
// double draw_result(int index, double value, double prev_value)
// void multiplex_use()
// void pass_forward()
// void pass_backwards()
// void multitude_initialization()
// void inputs_initialization()
// void outputs_writing()
// void weights_set()
// void weights_get()
// void weights_randomize()
// void set_text(int counter, double error=0)
// void create_text()
// void analysis()
// 
//+-----------------------------------------------------------------------------------------------------------------------+
 
And so on... I won't give you all the code
 

Roughly understood, Implex. After the nerve-racking, a link to an article like this... That doesn't sound like progress to me. If there had been anything thought-provoking in that publication, I would have understood it. But talk about how to squeeze a stat advantage out of a sports lotto drum, and further meaningful nonsense about how to beat Foreh, in my opinion, is unworthy of being discussed in such an interesting thread.

1...949596979899100101102103104105106
New comment
Reason: