neural network and inputs - page 5
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
Why are you sp'gashing? Are you looking for someone to blame for your illiteracy? DDD
Seriously, what's the catch?))
To me NS is one of the classification methods that only works in stationary markets. I ask the question as I have never received a reasoned answer.
PS. I do not care about illiteracy or illiteracy of the forum participants - I am personally interested in me, my personal favorite.
For me, NS is one of the classification methods that only works on fixed markets. I am asking the question because I have never received a substantiated answer.
PS. I don't care about the literacy or illiteracy of forum members - I'm interested in me personally, a very personal favourite of mine.
faa, do you understand what stationarity means?
faa, do you understand what stationary is?
You're still in my bathhouse.
So you don't understand....
And the stationarity of the process is not required to use neural networks for prediction. That's just in case.
For me, NS is a classification method that only works in fixed markets. I ask the question because I have never received a substantiated answer.
My opinion - NS does not like non-stationarity, due to the fact that it is an algorithm that requires a long learning curve. In other words, there is no such thing as a grid that reacts quickly to unforeseen changes in orculation. Unless, of course, the structure of the network itself contains a priori data about the nature of non-stationarity. But this is precisely the case where I would prefer NS to another, faster algorithm, as the beauty of NS is that it can be built without a priori data, hoping that the set itself will pick out the patterns. In other words the conclusion is the following: if we don't have any idea what model may underlying regularities (including non-stationarity), then we use NS, otherwise it's better to try and describe the model on paper, and then in matrix/differential/other form and take some good regression algorithm (or classification, clustering - depending on the problem). Although, by and large, the same algorithms can be used to train an NS built with theory in mind, so in principle it doesn't matter.
My opinion - NS doesn't like non-stationarity, for the reason that it is a long learning algorithm. In other words, there is no such thing as a grid that reacts quickly to unforeseen changes in orculation. Unless, of course, the structure of the network itself contains a priori data about the nature of non-stationarity. But this is precisely the case where I would prefer NS to another, faster algorithm, as the beauty of NS is that it can be built without a priori data, hoping that the set itself will pick out the patterns. In other words the conclusion is the following: if we don't have any idea what model may underlying regularities (including non-stationarity), then we use NS, otherwise it's better to try and describe the model on paper, and then in matrix/differential/other form and take some good regression algorithm (or classification, clustering - depending on the problem). Although, by and large, the same algorithms can also be used to train an NS built with theory in mind, so in principle it doesn't matter.
Stationarity is not required for NS.
A two-dimensional stationary process on a graph is a straight, horizontal line (or close to it). H.l. is not even required to be a NS to predict.
My opinion is that NS does not like non-stationarity, due to the fact that it is a long learning algorithm. In other words, there is no such thing as a grid that reacts quickly to unforeseen changes in orculation. Unless, of course, the structure of the network itself contains a priori data about the nature of non-stationarity. But this is precisely the case where I would prefer NS to another, faster algorithm, as the beauty of NS is that it can be built without a priori data, hoping that the set itself will pick out the patterns. In other words the conclusion is the following: if we don't have any idea what model may underlying regularities (including non-stationarity), then we use NS, otherwise it's better to try and describe the model on paper, and then in matrix/differential/other form and take some good regression algorithm (or classification, clustering - depending on the problem). Although, by and large, the same algorithms can be used to train an NS built with theory in mind, so in principle it doesn't matter.
My opinion is that NS does not like non-stationarity, due to the fact that it is a long learning algorithm. In other words, there is no such thing as a grid that reacts quickly to unexpected changes in orculation.
I wonder why use them then ? )))) Who here knows where the nets work in real time ?
People usually come to NS from frustration with technical analysis, on the one hand, and on the other hand, in an attempt to get, in fact, an indicator that no one else has. But all these people do not understand that the problem is in the non-stationarity of the market (variable mo and dispersion), and not in the absence of a miracle indicator, and if TS does not take into account this nuance, then any system, with or without NS, tends to rot and necessarily lose deposits. Therefore success on NS is possible just like success on TA. But TA is preferable because TA algorithms are much simpler, their internal structure is clearer and works much faster, different adaptation variants are possible. But still win on TA or NS - it is a win for an experienced player in the casino, no more.
PS. Many times wrote about it on this forum, just alsu confirmed my suspicions about the NS.
People usually come to NS from frustration with technical analysis, on the one hand, and on the other hand, in an attempt to get, in fact, an indicator that no one else has. But all these people do not understand that the problem is in the non-stationarity of the market (variable mo and dispersion), and not in the absence of a miracle indicator, and if TS does not take into account this nuance, then any system, with or without NS, tends to rot and necessarily lose deposits. Therefore success on NS is possible just like success on TA. But TA is preferable because TA algorithms are much simpler, their internal structure is clearer and works much faster, different adaptation variants are possible. But still win on TA or NS - it is a win for an experienced player in the casino, no more.
PS. Have written about this many times on this forum, just alsu confirmed my suspicions about NS.
that's it, that's the end...........
The end of TA and NS in the financial markets! Geez, geez....
P.s. faa, don't embarrass yourself.