Machine learning in trading: theory, models, practice and algo-trading - page 421
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 don't know, for my Target, I'm sure he's not peeking anywhere. Well, no, he does. But only into the future and not into the past. :-)
Very interesting - how?
How are the MA's?
I support the post above.
Following the thread I so understand the topic....
The historical data is analyzed for coincidence. Many definitions have been given here, including memory.
Drawing an analogy with typical strategies, a simple MA also has a memory. It basically results in some sort of averaging including by patterns.
This entire study boils down only to finding the most accurate market entry in terms of whether to buy or sell. But, as in any other forex strategy, there is no answer to the question of what to do if you are wrong. This question is the most important one, as no one is willing to give the trader his money so easily, no matter what system he has. The real quotes do not follow any laws and coincidences, they will go exclusively against the trader.
As I understand it, the possibility of an error cannot be excluded.
Very interesting - how is it?
How are the AIs?
I second the post above.
Following the thread I so understand the topic....
Historical data is being analyzed for coincidences. Many definitions have been given here, including memory.
Drawing an analogy with typical strategies, a simple MA also has a memory. It basically results in some sort of averaging including by patterns.
All this learning comes down to only finding the most accurate market entry in terms of whether to buy or sell. However, like any other strategy used in the forex market, there is no answer to the question - what are we going to do if we're wrong? This question is the most important one, as no one is willing to give the trader his money so easily, no matter what system he has. The real quotes do not follow any laws and coincidences, they will go exclusively against the trader.
The probability of error is not excluded, as I understand it?
Actually, it is very simple. If a signal shows a profit, I mark it with 1. If the signal shows a loss, I mark it with 0. Accordingly, the last signal, i.e. the current signal is considered undefined, because its outcome is not known. It will be known when the next signal appears. Well, the NS is trained to determine the signals in the current time, without looking into the future. So with the purity of data collection I'm fine...
Does anyone have any experience with mxnet api, bypassing R and Py? Alglib was not productive enough, does not work on GPU, complex networks take a long time. And in general a lot of productive libs on cpp
Does anyone have any experience with mxnet api, bypassing R and Py? Alglib was not productive enough, does not work on GPU, complex networks take a long time. In general, there are a lot of efficient libs on cpp
What is the difference - long or not, if you optimize (and you will have to optimize in any case), you can use all your CPUs/computers and cloud. And with cloud it will be faster than any other option - 256 parallel threads in genetics or up to 50-100 thousands (I don't know the limit) with full computation.....
A huge difference, you'll give a lot of money in the cloud with slow calculation of the NS, I had up to 7000 agents simultaneously in the genetic optimization, yes quickly, if the bot itself in the tester is run quickly, otherwise you're screwed... 15-20 minutes per core to train ns at 5k bars on i5 6200u skylake, and 2 more times retraining during testing. Better to rent a multi-core+gpu for this purpose and optimize without the cloud
+ diplerning if used is many times faster in itself, mxnet has
http://mxnet.io/
You strongly suggest that someone check. Check and give me the results.
Interesting to see and discuss the results.
Obviously I checked before making such statements that cross out the results of many local articles, on the application of MO in algotrading.
Once again I repeat Double-check the results on a random source, on random walk arithmetic or geometric. With ZZ and other fake targeting you will get prediction far beyond 50%, you can easily get 90%. You would have to prove that you can predict random, which is not reasonable.
I said something similar a couple hundred pages ago.
"Something similar" not that, you said about shifting ZZ, that it should\not be shifted by the bar, but it makes no sense, either way on SB would be grail, which is impossible. ZZ looks very far, it needs to be shifted by an undetermined number of bars, by at least one step between the knees.
All this learning comes down to just finding the most accurate entry into the market in terms of - buy or sell. However, as in any other strategy used in the forex market, still there is no answer to the question - what will we do if we are wrong? And this question is the most important in fact...........................
golden words..... everyone gets hung up usually on entry points..... the question of exit and accompaniment of positions in all the variety of scenarios is almost ignored... although in my opinion this is the most basic question, since the accuracy of entries is quite a mythical thing... the development of an action plan in all scenarios is important... after all this is what is completely in the power of the trader... something that can be fully controlled in contrast to market movements that cannot be controlled.....
Dear professors and associate professors of programming, have you finished the code?
Can I try it? At least a trial.