Machine learning in trading: theory, models, practice and algo-trading - page 2304
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
What I'm most excited about is the 'linear trend' label.
Well, his model is linear trend plus noise in the form of fractal Brownian motion.
Well, his model is linear trend plus noise in the form of fractal Brownian motion.
Then it should be nonlinear.)
Have you seen any new books on econometrics in R?
there are no such books on python, they're all in different places.
https://otexts.com/fpp2/
Nonlinear, then.)
Have any new books on econometrics in R appeared, have you seen them?
there are no such books on python, they are all in different places
https://otexts.com/fpp2/
econometrics is for chicks, DSP is for guys ))))
haven't readthis one? Although i doubt you'll find anything new
econometrics for chicks, DSP for men ))))
for tractor drivers
Then it should be non-linear.)
There, as usual, for each point we obtain a linear trend based on the history. Then this trend is prolonged by one unit into the future and a forecast is obtained. As a result, in my opinion, all the same, we will get a forecast in the form of a weighted average.)
have any new books on econometrics in R appeared, have you looked at them?
there are no such books on python, they are all in different places
https://otexts.com/fpp2/
I can't read such books, so I've stopped following them. For general theory I read high school or raech textbooks (Magnus, for example, or Kantorovich's lectures). For specific questions I consult manuals of necessary R packages, they contain everything, up to references to scientific articles in use.
econometrics is for chicks, DSP is for men ))))
haven't you readthis one? i don't think you'll find anything new.
as they say, you can read a book many times and always find something new.)
I'll read it. Econometrics + MO is all that is needed. But the grails don't lie on the surface. It is foolish to demand anything more from general collections.It's funny - a three-layer MLP network with 10-15 neurons in the middle layer found an x1/x2 type dependence.
The problem was the insufficient number of neurons - 8 for the middle layer is not enough
Ooh, the beeper is on, that's cool, if you don't turn it off in a week))
There are some features that, oddly enough, worsen the generalization ability (I speak for catbust in particular, probably applies to others as well). It would seem strange, because you just add new signs, and the model gives an error more than it was without them
for example, trained on several machines, then removed a few and the accuracy became higher
Why is it strange, as I understand it is a standard problem with all data. models are trying to take into account all the features, and if some have no relationship with the target mark, ie, are random, then it should be reduced quality.
Example: if you predict a person's weight from his height and sex, the quality is pretty high, but if you add hair color or some other crap to it, it goes down.
It's funny - a three-layer MLP network with 10-15 neurons in the middle layer found an x1/x2 type dependence.
The problem was the insufficient number of neurons - 8 for the middle layer is not enough
well, i told you so. and you want another joke - a network with 1000+ neurons - will not find the dependence or will be very long and inaccurate to learn.