From theory to practice - page 520

 
Олег avtomat:

formulas are not a mindless mass stampede. It's the path that led to the result -- the formula -- that counts.

well, if you get it in russian, i'll send it to you.

 
Maxim Dmitrievsky:

Well, if you come across it in Russian, I'll send it to you.

Thank you
 
Maxim Dmitrievsky:

Well, if I come across it in Russian, I'll send it to you.

I added it in there:

What he came up with there, I don't know. But I know from practice what the scope of polynomial regression is. Our (still Soviet) textbooks on computational methods say the same thing.


I will definitely have a look at the Russian ones.
 
Олег avtomat:

I added there:

What he arrived at there, I don't know. But I know from practice what the scope of polynomial regression is. Our (still Soviet) textbooks on computational methods say the same thing.


In Russian I will have a look at it.

I couldn't find a video, only an old book

Probably nothing special for those familiar with MoD, but maybe something new
 

There's only one video in Russian:

I couldn't get through it, but I think it's all relevant.


 
Maxim Dmitrievsky:

Couldn't find a video, just an old book.

The book is from 1979. (It's quite voluminous, I'll take my time reviewing it). However, I don't think there is anything drastically different from the contents of books by other authors published between 1980 and 1990 (e.g. Ivanov, Marchuk, Zeldovich, Myshkis, Samarsky, Gulin, Tsvetkov...).

But the video, as I understand it, presents them with some new developments. Right?

 
Олег avtomat:

The book is from 1979. (It's quite voluminous, I'll take my time reviewing it). However, I don't think there is anything drastically different from the contents of books by other authors published between 1980 and 1990 (e.g. Ivanov, Marchuk, Zeldovich, Myshkis, ...).

But the video, as I understand it, presents them with some new developments. Right?

I think this is old information too. From the new (last speeches) everything is in English.

specifically about bruteforcing search in I.O.D. as a form of artificial intelligence.

but the basis is the same - reconstruction of dependences by introducing mappings into new dimensions using kernel transformations and some other stuff on top

 
Maxim Dmitrievsky:

it is a normal idea to find a space in which they almost do not change, for example by bruteforcing, already discussed

strange, imho the law of change needs to be sought, well, BP cannot be characterised by a formula with constant coefficients, no matter what formula, even polynomial, even regression from Sultanov

For the second week I'm hooked on studying SSA-models, I'm interested in forecasting, moreover SSA model itself implies that there is a recurrence formula for the input series and it is sufficient to regroup the vectors of eigennumbers of covariance matrix....

I was studying MatLab codes about SSA, ported them from MatLab to MQL5 and watching your unhurried conversation I came to a conclusion, that the matrix vectors themselves should be forecasted, it's clear, you will get another "unreasonable forecast model" in the output, but it's not difficult to analyze matrix repeatability, with small sliding window, small matrices... i.e. reduce the problem to statistics

 
Igor Makanu:

strange, imho, the law of change should be sought, but BP cannot be characterized by a formula with constant coefficients, no matter what formula, even polynomial, even regression from Sultanov

For the second week I'm hooked on studying SSA-models, I'm interested in forecasting, moreover SSA model itself implies that there is a recurrence formula for the input series and it is sufficient to regroup the vectors of eigennumbers of covariance matrix....

I was studying MatLab codes about SSA, ported them from MatLab to MQL5 and watching your unhurried conversation I came to a conclusion, that the matrix vectors themselves should be forecasted, it's clear, you will get another "unreasonable forecast model" in the output, but it's not difficult to analyze matrix repeatability, with small sliding window, small matrices... i.e. reduce the problem to statistics

Well, yes, and preferably not to predict but to reconstruct the time dependence, for example, between 2 or more related BPs. There is no law there as it changes with changes in liquidity

 
Igor Makanu:

it is not difficult to analyse the repeatability of matrices, with a small sliding window, the matrices are small... i.e. reduce the problem to statistics

Alexander has the same problem, if there is no stable distribution, then identify patterns in their alternation, i.e. it all comes down to one thing, just in different ways.