Machine learning in trading: theory, models, practice and algo-trading - page 3000
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
Something like that? https://boostedml.com/2020/04/1-d-convolutional-neural-networks-for-time-series-basic-intuition.html
That's right.
a little tease with DSP " tease the stubborn ML-guys :-) and they are led and react amusingly
Although logically, all three things (DL/ML, DSP, and NN) should work together. The first one determines from history analysis what is noise/signal, the second one removes noise in real-time, the third one promptly highlights the signal. Or a live trader will have to be used instead of one of them.
The market sends you signals that are noisy? A forum of other topics will be useful for you.
Is the market sending you signals that are being jammed by someone else? The forum of other topics will be useful for you.
Do you want to provoke TC's coveted "ban for a month"? And you are naive...
That DSP methods can be used in data processing, and promisingly together with DL/NN, any objections ? except "my neighbour tried it, says it's not good".
Note that I just said that other methods should be used together with ML/DL. Each of them separately is unlikely to pull, but together it can be. And the result is like stepping on a sore callus :-)
PS/ which forum and topics I should read I will somehow decide myself, without your recommendations
You want to provoke TC's coveted "one-month ban"? You're naive.
That DSP methods can be used in data processing, and prospectively together with DL/NN, any objections ? except "my neighbour tried it, says it's not good".
Note that I just said that other methods should be used together with ML/DL. Each of them separately is unlikely to pull, but together can be. And the result - as if stepped on a sore callus :-)
PS/ which forum and topics I should read I will somehow decide myself, without your recommendations
You should first decide what exactly you want to process with DSP - either a signal with noise or some data. And it will be better if you decide in another thread - it will be the best proof that you want to discuss the topic and not just clutter this thread.
Why DSP is not suitable for price research I wrote recently, I don't see the point in repeating myself.
You should first decide what exactly you want to process with DSP - either a signal with noise or some data. And it will be better if you decide on it in another thread - it will be the best proof that you want to discuss the topic and not just clutter this thread.
Why DSP is not suitable for price research I wrote recently, I don't see the point in repeating myself.
What's the argument about?
Mocking any thread with mathematical methods is COC.
Also a clear sign of DSP is a return to historical data.
discretisation is again DSP.
and the funniest thing is that with all of the above, some "smart guys" are trying to measure my iq.
Stationary and quasi-stationary random processes. It is their realisations, from the point of view of mathematics, that are called signals. From the trader's point of view, it is an eternal flat, i.e. eternal trader's paradise with trading on the return to the average.
The difficulty in trading comes from the proximity of prices to the SB. The SB is NOT a signal. There is brown noise, it looks like a SB but it is NOT a SB.
I was looking through a couple of bourgeois bibles on COC. All of them at first write that "a signal is any sequence of numbers", but when it comes to mathematical specifics, it turns out that the signal is exactly what I wrote. For example, it is only for this case that you can define the ACF through convolution.
You mean this message? But it is just as desirable for MO to bring the data to a range. "Stationaryisation" would not hurt either. Instead of SB you can take increments or prepare the series in some other way. The book 10 pages ago about the use of spectrum analysis was written not by some amateur radio geek, but by a man who got a Nobel Prize for cointegration.
Maybe the requirements for MoD and DSP are not so different...
You mean this message? But for MO it is also desirable to bring the data to a range. "It is also desirable to rationalise the data. Instead of SB you can take increments or prepare the series in some other way. The book about the use of spectral analysis was written not by some amateur radio geek, but by a man who won a Nobel for cointegration.
Maybe the requirements for MoD and DSP are not so different...
Normalisation will not make the price series stationary because it does not change the ACF.
The increments are white noise that has no power spectrum (sometimes they say unlimited spectrum). Radio amateurs say their favourite mantra "any real signal has a limited spectrum" and instead of white noise they study some analogue of it with a trimmed spectrum.
The Nobelist's book, judging by the table of contents, deals with seasonal phenomena. On prices, such phenomena either do not exist or are obvious and inapplicable (daily volatility fluctuations, for example).
My main complaint about DSP is that it serves as mind-bait for novice traders with a mathematical background. Many of them start by applying Fourier decomposition to prices, quickly become frustrated with the results and give up trying. Instead, what is needed from the start is a thoughtful and much broader application of the whole of mathematics.
Economic time series is an extremely broad concept. Classical econometrics implies the presence of trend component, seasonal component, cyclical component and residual in the form of noise, the statistics of which is studied and modelled. That is, many economic time series are indeed a mixture of deterministic and random components. Accordingly, it is quite possible to identify the determinative component using DSP methods, the same spectral analysis, for example.
In the case of stock exchange quotes, and even more so in the case of forex quotes, this will not work, as they are close to the CB. I think even not too clever researchers will find the idea of analysing SB by spectral methods pointless.
Normalisation will not make the price series stationary as it does not change the ACF.
The increments are white noise, which has no power spectrum (sometimes they say unbounded spectrum). Radio amateurs say their favourite mantra "any real signal has a limited spectrum" and instead of white noise they study some analogue of it with a trimmed spectrum.
The Nobelist's book, judging by the table of contents, deals with seasonal phenomena. They either do not exist in prices or are obvious and inapplicable (daily volatility fluctuations, for example).
My main complaint about DSP is that it serves as mind-bait for novice traders with a mathematical background. Many of them start by applying Fourier decomposition to prices, quickly become frustrated with the results and give up trying. Instead, what is needed from the start is a thoughtful and much broader application of the whole of mathematics.