Does it make sense to switch from MT4 to MT5? Why did you switch to MT5? - page 16

New comment
 
Nikolai Semko:

I don't know about others, but I'm speaking from my own experience. On my algorithms, where there are millions of operations on each tick, it matters whether those millions of operations are calculated in 0.1 second or 0.01 second.

To have a 0.09 second head start is very cool in this frantic world of competition.

Still, I should probably clarify this phrase of mine, which for some reason has caused bewilderment among some forum participants.

There is nothing unbelievable about this figure.

A million operations is very little. Operations here, of course, mean all operations, including the simplest ones (comparison operations, sum, multiplication, etc.).

For example, I cite my ancient indicator, that I wrote more than 5 years ago in MQL4. It is quite primitive, because I just started to learn pattern recognition and didn't use OOP at that time. I'm even ashamed to refer to it.

In this unsophisticated indicator, I'm sure, there are more than a million operations on each tick, not counting the graphical part of the output. As on every tick all channels on all timeframes are recalculated.

 
Nikolai Semko:

I should probably clarify this phrase of mine, which for some reason has left some forum members perplexed.

There is nothing so unbelievable about this figure.

A million operations is very little. Operations here, of course, mean all operations, including the simplest ones (comparison operations, sum, multiplication, etc.).

For example, I cite my ancient indicator, that I wrote more than 5 years ago in MQL4. It is quite primitive, because I just started to learn pattern recognition and didn't use OOP at that time. I'm even ashamed to refer to it.

In this unsophisticated indicator, I'm sure, there are more than a million operations on each tick, not counting the graphical part of the output. As on every tick all channels on all timeframes are recalculated.

By pattern recognition, do you mean your regression indicators?
 
multiplicator:
By pattern recognition do you mean your regression indicators?
Well, you don't mean to say that the task of pattern recognition is just neural networks, do you?
Yes, this indicator uses linear regression to solve the channel recognition problem, but it is not a regression indicator.
I can assure you that you can recognise text even faster than neural networks with linear regression.
 
Nikolai Semko:
Well, you don't mean to say that the task of pattern recognition is all neural networks, do you?
Yes, this indicator uses linear regression to solve the channel recognition problem, but it is not a regression indicator.
I can assure you that linear regression can be used to recognise text even faster than neural networks.

What's so surprising about that. After all the fastest operations in computational processes are bitwise operations such as bitwise comparisons, shifts, summations etc.

One must take into account that a processor has a clock frequency of 4 GHz or more.

 
Yuriy Asaulenko:

Write now, (1) SanSanych has had a DLL for a long time, and it seems to be freely available. And all of the R ICL libs won't be able to handle it anyway. Otherwise it won't be a terminal, but a mixture of a bulldog and a rhinoceros.

(1) I have been aware of this for a long time. Just then, as I understand it, there will be an opportunity to flash a complex math processing with R functions in one instance (even if not all libs) and put it on the market.
 
Petros Shatakhtsyan:

What's so surprising about this. After all, in computing processes, the fastest operations are bitwise operations such as domestic comparisons, shifts, summations, etc.

One should take into account that a processor has a clock frequency of 4 GHz or more.

In modern multiscalar processors the mathematics of the clock frequency has long ceased to work.
After all, how many tens of elementary binary operations are needed to multiply two double numbers or to extract the square root from a double number? But rudimentary measurements on a 3 GHz processor indicate that multiplication of double numbers takes about 0.1( 10 Gigahertz per second) nanoseconds and extraction of the square root takes less than a nanosecond.
 
Nikolai Semko:
In modern multiscalar processors, the mathematics of clock frequency has long been out of order.
After all, how many tens of simple binary operations are needed to multiply two double numbers or to extract the square root from a double number? But rudimentary measurements on a 3 Hz processor say that multiplying double numbers takes about 0.1( 10 Gigahertz per second) nanoseconds and calculating the square root takes less than a nanosecond.

And this is from whom did you learn it doesn't work for a long time ?

But we have digressed from this topic.

 
Dmitry Fedoseev:

Propaganda for the great brakes. In time these atavisms like R and Python will die out like dinosaurs.

And only Fedoseyev will be left, hovering, blissfully above them all.

Python is a great high-level language, Go is coming as a replacement, also by Google, but for now Python is #1 for data processing and analysis.

You will get great brakes when you start working with tensors in low-level languages and very simple models, you'll be amazed at the amount of code you have to deal with.

 
Maxim Dmitrievsky:

And that will leave Fedoseyev alone, hovering, blissfully, over everyone else.

Python is a great high-level language, Go is coming as a replacement, also by Google, but for now Python is number 1 for data processing and analysis.

You'll get great brakes when you start working with tensors in low-level languages and simple models, you'll be blown away by the amount of code.

He won't have any problems. He won't.

 
Yuriy Asaulenko:

He won't have a problem. He won't.

That's why he's always talking nonsense in different threads )

By the way, if googlers move to GO in terms of MO, instead of python, speeds will increase significantly. And they will, sooner or later.

so it makes sense to start learning

1...91011121314151617181920
New comment
Reason: