From theory to practice - page 358
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
Okay. (chuckles) I'll investigate further. In about 2 weeks everything will become clear.
About TC on the basis of diffusion equations - if someone will like my deals (who wants - will find), having re-read works of Shelepin L.A. - ask questions, I will answer. But, distribution of model for VisSim is stopped and will not be renewed under any conditions.
So be it. I couldn't find any other predictive methods, alas... (the task was to find the work of a really outstanding person).
But, it turns out the only thing - only bringing BP to a stationary form can give results in neural networks. That's what I'm getting at. There is no other way and there cannot be. Right?
PS For general understanding of the current moment - TC based on diffusion equations has already been created, works and is no longer of interest to me. I want to switch to neural networks. Maybe this change of general reasoning on this thread causes misunderstanding?
(Ain't you tired of communicating with the local hamsters yet?) Soon they'll be rubbing their tights, sitting in front of the computer and foam at the mouth proving how bad you are, shitting themselves with the puppy delight to insert their stupid words and somehow express themselves
None of them have a coherent strategy or any results at all, and none of them seem to have much intelligence either. I got wind of that a long time ago. I wasn't that sincere on your part :)) I just hate to see you kicking back like a bear to a pack of these worthless cubs
Aren't you tired of communicating with the local hamsters yet?) soon they'll be rubbing their tights, sitting in front of the computer and foaming at the mouth proving how bad you are, shitting themselves with the puppy delight of inserting their stupid word and somehow prove themselves
and not one of them has a coherent theory or any results at all. I got wind of it a long time ago. I haven't been that sincere with them :)) It's just frustrating to watch you fend off a pack of these worthless cubs like a bear.
Tired. But, man, the forum is a bitch, it's got me... Can't escape :)))
So be it. I couldn't find any other predictive methods, alas... (the task was to find the work of a really outstanding person).
But, it turns out the only thing - only bringing BP to a stationary form can give results in neural networks. That's what I'm getting at. There is no other way and there cannot be. Right?
PS For general understanding of the current moment - TC based on diffusion equations has already been created, works and is no longer of interest to me. I want to switch to neural networks. Maybe this change of general framework of reasoning in this thread causes misunderstanding?
1) See, the thing is - people who have achieved outstanding financial (rather than mathematical) results in the market, do not publish their works, and in the forums mostly keep silent. Works of academic value are published, there is no money in them.
2) NS doesn't care about stationarity or non-stationarity, it is a universal approximator of anything.
"Bringing a series to stationarity" is a meaningless expression, the series is initially either stationary or not. It may be transformed to get another series, say, stationary. But what will happen after that? You trade the initial series, it means that inverse transformations must be done. For example, if you have forecasted the increments, to trade it, you will have to get a price series from the forecasted increments again by summation, and the forecast error will very quickly exceed all reasonable values.
It makes sense to try autoregressive models on H1 or D1 predicting one next increment that is already tradable due to its size. But not by tick increments predicting hours ahead as you are doing now.
3) Then where is the promised report on this TS? the signals only show the overall result for the month.
But, it turns out one thing - only bringing BP to a stationary form can produce results in neural networks. That's what I'm getting at. There is no other way and there cannot be. Right?
Neuronka is not stupid, it can trade on open price increments as well. Unfortunately the usual neuron at normal prices the profit will be lower than the spread, and that's bad. You need to use indicators (not wipers, but something more serious), or some kind of advanced deep neuronics, to squeeze more profit.
An alternative to indicators and deep nets is "to reduce BP to a stationary form" just as you wrote, on such BP even a neuronics with technologies of the last century will give quite good results.
Is it easier to reduce BP to a stationary form than to bother with indicators? I don't think so, there will be a lot of problems in any case, but it's a good idea.
Once again, look at your audcad ticks and the ones I sent you for comparison. There is a big difference between them, which Kolmogorov understands, but we don't. Somewhere in it is the key to everything.
Alexander has shown nothing but bluster and arrogance here) no wonder he's getting the same response)
Alexander has shown nothing but bluster and arrogance here) no wonder he's getting the same response)
Alexander is God here, he has invented and implemented the idea, and you are worthless schmuck without a name or a tribe, and there is nothing to call you.
You pounce when there are so many of you, but you have nothing to offer.
OK, I went a bit over the top somewhere - I repent.
I suggest we stop arguing for the time being - in 2 weeks we'll have another bout about stationarity and Erlang flows :)))
Did I do something to offend you?)