HOW to get a programmer 100% interested in writing an EA based on your IDEA - page 6

New comment
 

to YuraZ

I was like this once.
Called the head of the department and said - in two days to calculate the option directory in the sense of cost-effective
of 28 dyes for the dyeing room of a knitting factory. Here is the raw data, here is the equation.
Well I'm the equation let's gnaw and immediately opened unpleasant, it turned out that this simplex method with bilateral constraints, and direct solutions it is not.
I.e. the volume of my PhD dissertation and I need to give a working program for 2 days, the further into the woods the more reasons I got a Nobel prize for the simplex method.
But that's what programming is all about.
Two days later I bring a 40-line Fortran program, in which I loop 28 times on the number of optimized components.
The optimum mixture of colors by variants is calculated for 25 hours of machine time.
Then I explain 1 hour to my supervisor, and the supervisor explains a week to the customer. They have been expecting an agromodern programme from us)))
And we report in 40 lines. The customer gets nervous and resists the obvious.
And finally it turned out that the ministry had planned this task every year, the money was spent,
the money had been spent, but there was no guidebook from year to year - it's complicated by the simplex method.
So Lieutenant Rzhevsky is not a type at all, but a Hero.

 
SK. писал (а):

You don't have to work on someone else's TS.

Programming, as a field of activity, disciplines, aligns thoughts. This means that it allows you to screen out a large part of misconceptions.

The customer (to put it simply) is under an illusion, and the programmer is almost always aware of this. And taking money from him does... (qualify as he does). It is like the older kids at school playing cards.


SK, you see, besides the concept of MTS (mechanical trading system), that same magic chalk from the fairy tale, which is throwing pies and pancakes on the table, and which makes everyone go crazy, there is also the concept of ExpertAdvosor - an EA that helps a trader in his work. The programmer, being reasonable by definition, must adhere to realistic concepts (EA, as opposed to MTS). And it is not the task of the programmer, but of a psychologist to eliminate illusions that live in the client's mind. Who would have thought that these guys dream about a magic button which, once pressed, will turn their life into a magical wonderland!

If the concept of EA does not satisfy someone - please do not use them, sit staring at the monitor from morning till night.

 
Korey:

This is a simplex method with two-sided constraints, and there are no direct solutions to it.
I.e. the volume of a PhD thesis, ..., it is not for nothing that they awarded a Nobel Prize for the simplex method.

Nice job of rubbing your boss's points in.

But do not tell the mathematicians about it, otherwise they will faint :)

 
wenay:
NYROBA:
wenay:
I've found everywhere to be an enthusiast, but no one has even started to tell me their TS, they're afraid or ... =) I was just curious.

What prevents you from making your own profitable TS? ;)


it's no secret, I've already developed it, all that's left is for mts to write


I believe. ;) Why do I have to tell you twice, I'm not deaf. :))))
 
NYROBA:
I believe. ;) Why do I have to tell you twice, I'm not deaf. :))))

There was a glitch, didn't see my posts =)
 
Korey писал (а): Two days later, I bring in a 40-line Fortran programme with 28 times the number of components to be optimised in a loop.

The optimum paint mix by variant is calculated in 25 hours of machine time.

So... Alexander, tell me, on what supercomputer did you work, and in near-soviet time?

Here's the calculation. If the loop is nested 28, the minimum number of operations is about 2^28 (270 million), and if you take 3^28 (a more realistic estimate, though), then the output would be... 22,876,792,454,961 - in short, 23 trillion transactions. 25 hours is 90,000 seconds. Yeah, divided by each other, that's 254,186,583 ops. Wow, that's a nice computer you had back then...
 
Better:
Korey:

This is a simplex method with two-sided constraints, and there are no direct solutions to it.
I.e. the volume of a PhD thesis, ..., not for nothing they awarded a Nobel prize for the simplex method.

Nice job of rubbing your boss's points in.

But don't tell the mathematicians about it, otherwise they will faint :)


There was an artificial intelligence language called Prolog, it was a wonder, especially mathematicians liked it.
- You don't solve a problem in it, you just describe it, at the end of the description you say <goal>, and wait for the answer.
I had high hopes, but now Prolog is really under the chair - someone blabbed.
That Prolog's famed artificial intelligence searches for solutions with simple brute force, loop by loop.

I mean, you have to educate the customer until he understands what he really wants.

 
Mathemat:
Korey wrote (a): Two days later I bring a 40 line Fortran program with 28 times the number of components to be optimised in a loop.

The optimum mixing of colours by variants is calculated in 25 hours of machine time.

Taex... Alexander, confess, on what supercomputer did you work, and at a time close to the Soviet one?

Here's the calculation. If the loop is nested 28, the minimum number of operations is about 2^28 (270 million), and if you take 3^28 (a more realistic estimate, though), then the output would be... 22,876,792,454,961 - in short, 23 trillion transactions. 25 hours is 90,000 seconds. Yeah, divided by each other, that's 254,186,583 ops. Wow, that's a nice computer you had back then.

That was five years before KinDZA came out.
Fortran was a novelty. There were no software packages. Computers came barebones, even without DOS, in particular the first Eescs.
I might have made a mistake about 28 colours. I remember only that the first reference book had 28 pages.
It's like maths: if you don't practice, you forget, so I forgot.
And I explained to my boss - that for the allotted two working days it was only possible to write and debug 40 punched cards.
-that there was a simplex method available with a one-way restriction,
and we have bilateral ones, which the science of the time hadn't yet chewed up and read as a scientific problem.
Now - yes, everything is ready-made. even drag and drop into MQL-4.
 
Korey: There was an artificial intelligence language called Prolog, it was considered a miracle, and mathematicians especially liked it - you don't solve a problem in it, you just describe it, give <goal> at the end of the description, and wait for the answer.
Yeah, Trubo himself taught Prolog to fucking 5th year students. It's a curious language, like AI. Very easy to write recursive calculations of determinant in it (literally couple dozens of lines), but bastard, seems to count more than 6th-7th order with great difficulty. Here it is, AI with great promise...

P.S. Well, I wouldn't say that it's so perfect; you just describe it, and the problem solves itself. This is only in an ideal, for a completely non-procedural language. The main problem of this language is an avalanche-like increase in the number of variables due to recursive calculations. This makes it rather a nice theoretical toy than a language for AI tasks. But it's a pretty good exercise for the brain, especially after procedural languages like Pascual or C.
 

to Mathemat

By the way, I have already come across ideas of Expert Advisors with blunt search for various indicators important for this market and maybe with search of various TS.
I.e. the idea of searching for <goal> options may turn out to be relevant.
After all that Prolog was used on very slow computers, now it is 2 orders of magnitude faster,
and memory is planar now, not like before - in 64K pages.

12345678910111213
New comment