Searching for an arbitrary pattern using a neural network - page 7

 
Реter Konow:

Do you think that NS is such a "magic wand", that whatever you give it, you always get what you need? No matter what kind of data, no matter how big it is, it is all the same - the numbers for NS.

Then I do not understand, where is the algorithm that finds all the possible patterns? Where is this "almighty" NS? They've been studying MO for so long and still no "pattern recognizer" in MT's arsenal.

not in my opinion, but the NS is an algorithm, the fact that this algorithm is called the NS... well, it is necessary for the industry, the main problem is just in the preparation of data - they are literally prepared by hand, or almost by hand

ZS: There are prediction systems, there are self-learning algorithms... Look on youtube about Tesla, you'll get a lot - there's information about advanced recognition technologies - if you don't want to read it, but I suspect that you'll end up on a popular video, where everything is sort of intelligent and not done by engineers ))))

 
Igor Makanu:

not in my opinion, but NS is an algorithm, the fact that this algorithm is called NS... Well, for the industry, the main problem is in data preparation - they are literally prepared by hand, or almost by hand.

So that's what it's all about. The result depends on the data. And here the data is fundamentally different, both in type, volume and content. Maybe, no, definitely, it should influence the result.
 

Peter, in general I do not want to get into your spatial reasoning, I remember the topic about OOP, you rarely allow yourself to read primary sources, and without the matrix communication with you will look like I will once again fight with windmills - here I am tired of it, with all due respect to unfamiliar people.... there's a lot of profanity going on here.

)))

 
Igor Makanu:

...

ZS: there are prediction systems, there are self-learning algorithms... But all the same it's a work with numbers based on an algorithm, there's also a database, but they still collect data mostly by hand, look at youtube about Tesla, you'll get a lot - there's information about advanced recognition technologies - if you don't want to read, but I suspect that you'll get to the popular video, where everything is sort of intelligent and not engineers were doing )))

I'll have a look. It's interesting. But I read your article, and it clearly separates the application areas of networks. Classification, Prediction, Recognition. We are talking about recognition, and therefore the data must have a "visual" character. Well, it is at least logical.

 
Реter Konow:
Can you as an expert make an NS that recognises at least 5 patterns on any chart and timeframe?

Which chart and which timeframe matters not at all. 5 patterns is on the task, networks used to recognise whole alphabets back in the last century.

 
Igor Makanu:

Peter, in general I do not want to get into your spatial reasoning, I remember the topic about OOP, you rarely allow yourself to read primary sources, and without the matrix communication with you will look like I will once again fight with windmills - here I am tired of it, with all due respect to unfamiliar people.... there's a lot of profanity going on here.

)))

So that's the mate that would explain it. I would have accepted. Otherwise, laughing, generalities... All right, thanks and that's it.
 
Dmitry Fedoseev:

Which chart and which timeframe matters not at all. 5 patterns is on the task, networks used to recognise whole alphabets back in the last century.

Your NS must have been recognising all the patterns for a long time now.
 
Реter Konow:
So that's the mate part and it would be explained. I would have accepted it. Otherwise, laugh, general words... Ok, thanks and that's it.

I do not know how to teach, the links - yes all google, hobr you have already found, there are articles on the NS from the level of a pure zero, to the level of a pro

but download any book, as I wrote above - any next book on the NS more than half will repeat the first, alas this is how the explanation of the NS material - the essence is quite small, mostly it comes down to what type of NS to use and the preparation of data

 
Реter Konow:
That's what I wrote, that's what he does. Consistently identifies shapes by scaling the focus of the gaze. By the way, one operates with information in the same way. Consistently abstracting and detailing meaning.

No, that's not it at all. A man picks out the main thing. Something that sticks out.

 
Реter Konow:
You must have had the NS recognise all the patterns a long time ago.

No. I recognise them in other ways.

Reason: