![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
No.
They just try to do moves that are not possible in the beginning, which result in a immediate losing of the game by disqualification.
Then they will discover and learn what piece can do what moves, automatically over time.
It does not know the rules when it starts but since an illegal move results in a direct loss it will teach itself not to make those again.
This is what it does by itself, you do not have to teach it anything.
At first it will just place random orders and observe the outcomes.
But then it will have placed so many orders that it will start to identify patterns.
This results in pattern recognition and the sooner it will be able to tell what pattern is evolving, the sooner it can enter the trades.
This is no different then voice or facial recognition or reading human handwriting or recognizing a cat in a picture.
These techniques usually start out by using a base neuronet that is universally adaptable and expandable in size.
Then the input environment and gates are set up and configured and the network is trained resulting in the weights output for each neuron.
Yes raw tick data feeds.
I just mentioned some examples these techniques are already being used widespread.
The network only has to be adapted to learn or train on trading financial markets.
Your best start would be to look on youtube for terms like machine learning and AI pattern recognition there is a large quantity of information available.
I would certainly recommend you to visit the Tensorflow channel.
No.
They just try to do moves that are not possible in the beginning, which result in a immediate losing of the game by disqualification.
Then they will discover and learn what piece can do what moves, automatically over time.
It does not know the rules when it starts but since an illegal move results in a direct loss it will teach itself not to make those again.
This is what it does by itself, you do not have to teach it anything.
At first it will just place random orders and observe the outcomes.
But then it will have placed so many orders that it will start to identify patterns.
This results in pattern recognition and the sooner it will be able to tell what pattern is evolving, the sooner it can enter the trades.
This is no different then voice or facial recognition or reading human handwriting or recognizing a cat in a picture.
These techniques usually start out by using a base neuronet that is universally adaptable and expandable in size.
Then the input environment and gates are set up and configured and the network is trained resulting in the weights output for each neuron.
But then it will have placed so many orders that it will start to identify patterns.
habibo, "so many orders", there is a cost for that.
edit : you add that markets conditions aren't fixed either and you have a big problem of feasibility particularly if you reason on the tick basis (disk space for the sample to train & mem for training millions ticks)
But then it will have placed so many orders that it will start to identify patterns.
habibo, "so many orders", there is a cost for that.
edit : you add that markets conditions aren't fixed either and you have a big problem of feasibility particularly if you reason on the tick basis (disk space for the sample to train & mem for training millions ticks)
That doesn't even have to be ticks even M1 open data can work.
It's trained on historic data how does this cost ?
It's not necessary a high frequency project these networks are extremely malleable.
I am currently running two simulations which use a 32 Kilobyte sample size, i thought a Commodore64 had 64 Kilobyte that's twice as much.
I would not start like you by describing or trying to come up with supposed big problems before even trying yourself, because i have already done some work in the past and it was a piece of cake.
I didn't have any problem at all it's just that the training takes a decent bit of time.
And every time you find improvements you have to re-do the training so it's a time consuming activity.
That doesn't even have to be ticks even M1 open data can work.
It's trained on historic data how does this cost ?
It's not necessary a high frequency project these networks are extremely malleable.
I am currently running two simulations which use a 32 Kilobyte sample size, i thought a Commodore64 had 64 Kilobyte that's twice as much.
I would not start like you by describing or trying to come up with supposed big problems before even trying yourself, because i have already done some work in the past and it was a piece of cake.
I didn't have any problem at all it's just that the training takes a decent bit of time.
And every time you find improvements you have to re-do the training so it's a time consuming activity.
And how was the outcome of those past works?
This robot was not trained to trade but it was fed historic data and then was told to draw a graph forward in time, and to 'predict the future' based on the data it was given.
This robot was not trained to trade but it was fed historic data and then was told to draw a graph forward in time, and to 'predict the future' based on the data it was given.
Well it wasn't really useful because the bandwidth it came up with was still way too wide to make up a definite direction.
But this particular project was back in 2016 and things have since improved rather dramatically.
Still i would vote that trying to predict the future is not the right approach, since it is impossible.(and most likely unnecessary)
Well it wasn't really useful because the bandwidth it came up with was still way too wide to make up a definite direction.
But this particular project was back in 2016 and things have since improved rather dramatically.
Still i would vote that trying to predict the future is not the right approach, since it is impossible.(and most likely unnecessary)
Very interesting. If we are speaking in terms of candles, how many bars should be included for a EA of this kind to work? the last 24 candles? 48? 100?
That doesn't even have to be ticks even M1 open data can work.
It's trained on historic data how does this cost ?
It's not necessary a high frequency project these networks are extremely malleable.
I am currently running two simulations which use a 32 Kilobyte sample size, i thought a Commodore64 had 64 Kilobyte that's twice as much.
I would not start like you by describing or trying to come up with supposed big problems before even trying yourself, because i have already done some work in the past and it was a piece of cake.
I didn't have any problem at all it's just that the training takes a decent bit of time.
And every time you find improvements you have to re-do the training so it's a time consuming activity.
Then respect for that dude, because it's an arduous project you have there. I'll be glad to be informed of the results you got by doing it that way because it's an extremely purist approach.
Very interesting. If we are speaking in terms of candles, how many bars should be included for a EA of this kind to work? the last 24 candles? 48? 100?
The idea is that it would analyze the last 100 candles OHLC and find patterns by experimenting and learning thru many many (many many many ...) deals which ones suits the market. There's probably a system of reward somehere also.
I understood well, did I ?