Machine learning in trading: theory, models, practice and algo-trading - page 3350

New comment
 
fxsaber #:

On the same list.

I remember somewhere you compared max. profit among dts. And on a particular chart, what algorithm was used to obtain max. profit? Through optimisation or is there a strict algorithm.
 

The methodology of conformal predictions also echoes kozul, at least in terms of inverse probability weighting. I haven't read further yet. A lot of definitions :)

And the definition of potential outcomes is used in the same way. But it is already more clear for the case of binary classification. That is, no tritment or instrumental variable is introduced.

GitHub - valeman/awesome-conformal-prediction: A professionally curated list of awesome Conformal Prediction videos, tutorials, books, papers, PhD and MSc theses, articles and open-source libraries.
GitHub - valeman/awesome-conformal-prediction: A professionally curated list of awesome Conformal Prediction videos, tutorials, books, papers, PhD and MSc theses, articles and open-source libraries.
  • valeman
  • github.com
A professionally curated list of awesome Conformal Prediction videos, tutorials, books, papers, PhD and MSc theses, articles and open-source libraries. - GitHub - valeman/awesome-conformal-predicti...
 
My last 2 articles, at a simple level and without nuance, pretty much describe all of these approaches. Let's say they don't describe them, but they come close. I'm now checking the details of what they have researched. For example, inductive from transductive conformality differs only by one or two classifiers, separately for each class label. The latter is better (more accurate) at estimating posterior. And I used the inductive method. Another thing there is to retrain the models with adding and discarding each sample, for more accurate estimation. It's very expensive, but kind of efficient. But you can take simple and fast classifiers. Which I also wrote about while training on stumps.

I don't see any applause for my brilliance



 

Hi!

I'm trying different ways.

And the NN+GA algorithm is paying off. Much more stable.

 
Alexander Ivanov #:

Hi!

I'm trying different ways.

And the NN+GA algorithm is paying off. Much more stable.

Are you saying you're cooler than me?
 

an eveningread with vodka, venison and cucumber.

developing the theme and trying to link in my head approaches from different MOSH disciplines.

A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification
A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification
  • arxiv.org
Black-box machine learning models are now routinely used in high-risk settings, like medical diagnostics, which demand uncertainty quantification to avoid consequential model failures. Conformal prediction is a user-friendly paradigm for creating statistically rigorous uncertainty sets/intervals for the predictions of such models. Critically, the sets are valid in a distribution-free sense: they possess explicit, non-asymptotic guarantees even without distributional assumptions or model assumptions. One can use conformal prediction with any pre-trained model, such as a neural network, to produce sets that are guaranteed to contain the ground truth with a user-specified probability, such as 90%. It is easy-to-understand, easy-to-use, and general, applying naturally to problems arising in the fields of computer vision, natural language processing, deep reinforcement learning, and so on. This hands-on introduction is aimed to provide the reader a working understanding of conformal...
 
Maxim Dmitrievsky #:

aread for an evening of vodka, venison and cucumber.

developing the theme and trying to link in my head approaches from different MOSH disciplines.

Bon appetit and a mild hangover)

It seems to be very similar to probabilistic forecasting, although they write that they are different things. As far as I have understood so far, conformal is more focused on classification, and probabilistic is focused on regression.

 
Maxim Dmitrievsky #:
I remember somewhere you compared max. profit among dts. And on a particular chart, what algorithm did you use to get the max profit? Through optimisation or is there a strict algorithm.

And one-pass. Somewhere on the forum.

 
Aleksey Nikolayev #:

Enjoy your meal and have a mild hangover)

It seems to be very similar to probabilistic forecasting, although they write that they are different things. As far as I have understood so far, conformal is more specialised for classification, and probabilistic for regression.

Thanks :) yes, similar. They write that it doesn't matter classification or regression. How to get estimates for predictions via comparison on the validation network is clear (in the case of "Inductive", i.e. faster and simpler way). "Transductive" is also more or less clear, but it is very slow because it requires training as many models as there are examples in the sample. There are also intermediate variants like CV, which I actually did myself.

I didn't quite understand from the article how the final models are trained, what is substituted where. Again through correction of model weights, its calibration (sample weighting) or something, like in kozula. Or the most probable markers are substituted into the model after evaluation. I just used the second model for this purpose, which prohibits trading on bad examples.

 
Maxim Dmitrievsky #:

aread for an evening of vodka, venison and cucumber.

developing the theme and trying to link in my head approaches from different MOSH disciplines.

for medicine.

where the graphs crawl between two parallel lines,

which is nothing compared to the financial markets.

---

I smoked the gradient descent over the weekend.

You can do it without the I.O.D. in a heartbeat.

i.e. approaching the extremum:

x0-x1

x0-x2

x0-x3

etc.

There's something to that, of course.

1...334333443345334633473348334933503351335233533354335533563357...3700
New comment