I fully consider this resource one of the most powerful in the Runet for the rabble of idlers and labotryas, but at the same time specialists of the highest level in the field of defense. One does not interfere with the other as they say 🙂
The regulars know that I recently got a job and I was faced with the fact that the company just might need MO technology. And they need it already badly, but since they have never had a person with my experience, then they think that there is such a tool that, in principle, is able to do the work of choice as well as it is done by a person with his to (AI) pedantry to accuracy and speed of task completion. The person here will never be the first. Whatever you say. machines are now much dumber than we are, but they count much faster and more accurately. What is this branch actually for???? My fans who support me constantly the same Maksimka, Vizard (I noticed how much you respect druzhe :-)) Renat, FilPanAleksey Panfilov, NechaevRealleAlexander Nechaev, 7Konstantin7Anatolij Anufriev, AndreySAndriy Sydoruk, DNGDmitriy Gizlyk, geninoEvgeny Belyaev, sithoryHatem Koshok, msolito59Marco Solito, NovajaNovaja, shekishlekiPavel Bobrovskiy, seliveruSEM, moneystrategyVladimir Gribachev, VOLDEMARVladimir Pastushak, KirillovYVYury Kirillov, 80Viktor80Viktor, awan-fxAgus Setyawan, GaintAleksandr Gaint, KuripkoAleksandr Kuripko, AlPrasAleksandr Praslov, alexdomAleksej Kravcenko, le107_Aleksey Altukhov, terentyev23Aleksey Terentev, alekseyka33Aleksey Yagneshko, Razvan12Alexandru Razvan Gheorghe, alfaforexAlfa Forex alextrade25Aliaksandr Maksimau, FXboxINNAliaksei Tsestau, DejavuzAllan Shen, Anastasiya_1987Anastasiya Astakhova, old_padreAndrey Kolmogorov, yaav89Andrey Yakovlev, degalyuk97Andrii Dehaliuk, Anton078Anton Sidorchuk, Azamatik1992Azamat Sarsembaev, MilonegBogdan Yarotskiy, MaMonChonladhon Chittapakob, FAVORITE_XDenis Kotsubinskyi, DenizripDenis Pershin, thewhisperDmitriy Piskarev, Dima_SDmitriy Skub, ddpolkovnikDmitry Pilshtshikov, Dr. TraderDr. Trader, timemanEvgeniy Sklyarov, PammClabEvgeniy Yushkov, XpucmocEvgeny Proskurnya, virus3Evgeny Raspaev, forex777winFXwin, FreimansFreimans, TRaiderGeorgy Pankrashkin, MakatiBanditGraham Heyes, 603912Harun Celik, edutakIbragim Dzhanaev, igorbayt552igor baytuganov, pepsiezjirayut kemajaru, kostik83konstantin Kotlov, podgorniykvkonstantin podgorniy, mhfx7mani heshmat, masminmarcus Kaehler, mishutka999mikhail Mitin, Mostafa.sabrimostafa Mohamed Sabri, bermaui314muhammad elbermawi, modsellermustafa Eroglu, natalinvnatali Honchar, kinmaxnatalja romancheva, niktronnikolay Gaylis, alehfxoleg Bakurov, elenaviktorovaolena Kondratenko, missandraolena zelenska, lisyanolesya Galyamova, ypypavel Izosimov – IPA investments Ltd, rapridepavel Predein, roshrashid Umarov, eyvazovr1991rustam eivazov, specoolantserge, digimaxsergey don, sergoalsergey Kutsko, lotusussergey Ukhobotov, astrendsergey Zhukov – as trend LLC, ser1970sergio garziera, tradeez-fxtibor rituper, budriganuladzimir budryk, uladuladzimir Izerski, perepelvasily Perepelkin, gtkapitalvitalie Turcanu, ouyangximengxiaoneng Meng, Goldenrate20yaghoub Karampour, robot_hft_fp_v1yaroslav Yatsenko, Reshetovyury Reshetov, Capursiumcapursium, forexman77forexman77, mader11sendkod, Toxictoxic, masterlg ‘ Ґajґ© “® know that I am an adherent of the trenar classification Which Will not be enough to solve the production problem and that is Why the creation of this Branch Took place.
Within a few posts, I will form a statement of the problem and ask you to recommend the direction and relevant literature, that is, where you can look in order to solve the problem so necessary for production. I will say right away that initially there are from 10 to 50 categories for which you need to classify or rank, to be precise. The area under study is VERY large with a bunch of not obvious solutions, but this area is finite, which gives hope that it is AI that will be able to cut it. The area is large, finite, and most importantly static, which is why I have high hopes for AI. But the task here is completely different. It’s not up or down… here also in the side and straight and with a perepodvypodvert… well, You know what I mean. Especially I ask not to litter the topic…….. allow time for formulirovki tasks. Thanks!!!!
unicornis
In the beginning, you need to make sure that the devices are not clones of bourgeois / Chinese models. Look at analogs in the Internet, if the handle buttons are similar-100% clone. Promote a couple of copies, see what is written on the boards(maybe what was not scraped from the inscriptions), what small circuits, look at analogs in the Internet. It is necessary to start with the main and big idea of development, the development budget, its coordination, work plans, etc. Development from scratch and the release of devices is very intellectually expensive, there is no element base in the CIS, so 99.99% is China with names in the national language and localized firmware. (According to all requirements, the device must be domestic, so a box with the necessary letters is ordered and the firmware is localized from the finished solution).
vitales
I think so, if I understand you correctly. You need to assemble the device from the available set of components and at the same time minimize the cost of parts and that the specified restrictions would be fulfilled. Then you have the usual task of finding the optimal one. There is a good book on this topic – Taha Hemdi A. Introduction to operations research.
virus3
I’m leaning towards to separate NS. 1 NS classifies models by features. 2 NS classifies models based on the input query. But intuitively I feel that something can go wrong. I think the models should be divided into submodels.
Model 1.1, Model 1.2, Model 1.3 – this way I think you can avoid ” forks”
yuba
A properly made database will produce all this based on the usual statistics. Marketing tasks from time immemorial were successfully solved by collecting and stat. processing of information. I don’t think your task is unique.
Now, by the way, I do just such a database, for similar tasks, but in a completely different area. I think that the tasks are much more complex – there are thousands of different parameters and interacting objects. Without a database, it is impossible to cover all this. I can’t say the area, it has nothing to do with the market-trade-Finance. By the way, my task is also not unique.)
genino
There are no such cars. You need to buy 2 cars. One for my mother, one for myself. If you need to find a car that is perfect for the buyer, you need to know the needs of the buyer. In General, they often choose from 2-4 models and already know about what they will buy. Buying a car is not buying a TV. The campaign does not need ML for these tasks.
Now, if Pribluda along the way to sell mats, radio, heated seats… something else might come in handy.
nikelodeon
I agree, but if the request contains the values of the selected parameters. But the request includes hardware requirements. As an example: you Need a car to drive on the races and mom to safely go shopping. What should be done in this case? Only AI I think so…
virus3
The problem of classification comes to mind only such a model. All converge on the classification of the model. Input the Incoming request and model elements are received. I think like that. I don’t remember the type of network if I find it, I’ll write it.
yuba
There is no need for MO, but a regular database, and regular queries to it.
nikelodeon
So let’s continue……. I Remind you that I need help in choosing a method and approach. You don’t need to do anything for me, just tell me what to read.
The essence of the problem: the Input request has a set of technical parameters on the basis of which the choice is made.
The company sells “Device” equipment. The “device” has about 10 models of modifications that give it different properties. Each of the models has its own specific set of internal parts, where each of the spare parts can be made of different materials. Let’s assume that the incoming request assumes one and only solution. The structure is very similar not to the forest, but for solving it to use a random forest, I think it is unlikely to work. Let’s take a look at the diagram….. Immediately I apologize was done hastily the task was to show the approximate structure.
It seems to be nothing complicated, even the NS idea is not needed here. BUT the fact is that the area itself is too large. And models of order 50 and each model element of order 10. But the problem is that there is no complete number of possible input queries and there are such queries where the solution will be one that was not in the training. Plus there is always a fork when choosing when two or three models fit at the same time. One way or another, the GURUS of MO listed above, including the blue color, tell us which method will be relevant here. Exclusively approach. Who would use what here. Thanks!!!