https:\/\/decoders-zolotarev.ru\/en\/books\/<\/a> , where very different opinions are expressed about our book, published in the anniversary year of Shannon’s great article on codes. Look at and evaluate the level, style, and issues discussed in those comments. This is very instructive and very useful. It is clear they are in Russian.<\/p>\nSo one of the commentators of our achievements, the review of which on our book, perhaps we will also be able to show, proposed a very tough comparison of all published algorithms. He rightly pointed to the need for calibration programs for decoding algorithms for one or another, and better \u2013 for all the main channels <\/u>considered in the coding theory.<\/p>\n
It is not surprising that as such programs it was proposed to use our highly reliable software models of MTD decoders and communication channels in the C++ language, which guarantees the highest correctness of comparison of various new decoding methods with our algorithms, in many cases already really and for a long time working near the Shannon’s boundary.<\/p>\n
We have clearly realized the highest value of our wise colleague’s proposal. This approach after the introduction of mandatory practice in the development of decoding algorithms will be able to significantly reduce the possibility of many errors in the proposed publications about the new decoders and the apparent deception regarding their capabilities.<\/p>\n
So, what is a calibration program based on MTD algorithms?<\/p>\n
We have prepared several types of MTD decoders demo programs based on algorithms for convolutional codes, which contain ~10 threshold elements (10 decoding iterations). In accordance with the decisions taken to save the time of preparation of such demo programs, they were taken as simplified and reduced by 5 \u00f7 50 times versions of our research programs. Since the computer mainly simulates the operation of threshold elements (TE), then, determining the time of operation of the MTD on such a computer, we get its speed. For example, if the code distance d for codes that are used in such an MTD calibrator, when this distance is also close to d~10, at 10 iterations we obtain the computational cost of the computer, equal to N0<\/sub>=I0<\/sub>*d0<\/sub>~ 100 operations of adding small integers for each decoded symbol. Since for such the most preliminary estimates of the complexity of the MTD can not take into account that sometimes the MTD compares and changes the decoded characters, the proposed assessment is quite acceptable, since in all our programs-calibrators, we specify either the working time of the program or immediately print for you the speed of the MTD. Note that the calibration programs are needed as reference points to indicate the speed of the computer on which the new authors will show the capabilities of their programs for decoders.<\/p>\nThe model of the algorithm, which is presented in the C++ language by the applicant or just the author of the article, should demonstrate effective work on the computer under the conditions that the author considers the most advantageous for his method, but choose, nevertheless, the typical combination of parameters, i.e., for example, the code rate R=1\/2, etc. But in order to compare simultaneously the new method with some MTD decoder in typical decoding techniques and the same other conditions, only one more operation is needed. Our methods are characterized by the fact that we in publications usually specify both the number of decoding iterations I, and the code distance d. In addition, we often write that, for example, we used a laptop with the Core-i7 processor. So, if you choose the MTD decoder working in the same channel as the new algorithm proposed by someone, with other products of N=I*d than the calibrator, then the difference between N and N0<\/sub> on the same computer will indicate that the real MTD will be in N\/N0<\/sub> less fast than the calibrator. We believe that in fact, to test new decoders the problem will be to find such parameters of the channels that the new decoder can withstand these tests, as well as to be able to convince readers that the real presented program provides exactly the parameters declared by the author of the article or thesis. That’s all!<\/u><\/strong><\/p>\n But even such a simplest filter for decoding algorithms will be very strict for the authors of new error correction methods. We suggest everyone to start working with the applied problems of coding theory, finally, quite correctly.<\/p>\n
We, for our part, will try on the first steps of forming a bank of real effective new decoders, the correctness of which can necessarily be strictly checked (alas, otherwise it is impossible(!), now there are everywhere the sea of hack-work and just frank deception), to help new authors to coordinate parameters of their methods and characteristics of MTD decoders (by the way, sometimes it is necessary to involve in comparison VA and our block VA!). This will really serve the development of applied coding theory in all its most important aspects.<\/p>\n
Of course, the MTD-demoprograms spent some resources of laptop for generation of noise, transmitted data, etc. But the share of costs in all our programs are always small. These questions can be clarified in the future with the development of our proposed methods of comparison and calibration of algorithms.<\/p>\n
In the meantime, we’re launching the first version of the calibrator bank.<\/p>\n
Well, now about these programs.<\/p>\n
So far we have prepared 4 programs: for symbolic codes, for erasing channels, as well as 2 programs for the Gaussian channel. The first of these last two calibrators uses an accelerated version of the MTD decoder described in our previous monographs and in the Russian Handbook, and the second uses the basic conventional version of this algorithm as it is usually implemented as a software. Each calibrator in the instructions inside their libraries is described very simply and uniformly, to read the necessary information about them you do not need a lot of time. The descriptions themselves are in each of the four folders with instructions and calibrators that are packed together. Once again, the calibrators are used for a simple assessment of the complexity (decoding speed) of new algorithms when modeling them in C++ and comparing them with MTD algorithms on any Windows laptops for standard, but the best conditions in which their authors are ready to show their advantages. A link to calibrators is given on the «Education» page of our portal at the end, i.e. at the bottom of its right column. The total size of the four folders in compressed form is quite small and overwriting four calibrators does not create problems. Use them all!<\/u><\/strong><\/p>\n Some parameters of calibrators, as you will see, can be changed a little. But this is not their very important property. Useful for the first experiments with MTD algorithms need to use our demoprograms and software platform for different codes and channels available at the portal, mainly, on the page «Education». These platforms have much more opportunities to work with MTD decoders in the first steps of studying and even research OT and MTD. They are located on «Education» page of our portal.<\/p>\n
So to the true pioneers in the field of the best new real and concrete decoding algorithms \u2013 our greetings and unconditional all-round support!<\/u><\/p>\n
\n
Write. Consult. Support<\/u>.<\/p>\n
\n
Scientific school OT<\/u><\/strong><\/p>\n","protected":false},"excerpt":{"rendered":" Dear colleagues! We together with you can be very pleased with how successful was for the theory of error-correcting coding anniversary year of the great article by C. Shannon «Mathematical theory of communication.» On the example of about a hundred very successful development of decoding algorithms in recent decades on the basis of a wide […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[1],"tags":[],"yoast_head":"\n