RU2797406C1 - Random series generator - Google Patents

Abstract

FIELD: computing technology.

SUBSTANCE: random series generator consists of the first selector multiplexer 1, the second selector multiplexer 2, the first register 3, the source of random numbers 4, random access memory 5, K≥2 storage units of interval boundaries 6₁-6_K, K comparison units 7₁ -7_K, priority encoder 8, N≥1 inverters 9₁-9_N, the second register 10 and the neuro-fuzzy network unit 11.

EFFECT: increasing of the reliability of the generated series.

2 cl, 4 dwg

Description

The invention relates to computer technology and is intended to generate a random sequence of values from a given set of values with the required characteristics of the generated sequence.

A random sequence generator is known according to RF Patent No. 2250489 "Random Sequence Generator" IPC G06F 7/58, H03K 3/84 published on 20.04.2005 Bull. No. 11, including a source of random numbers, an N-bit selector-multiplexer, random access memory, an interval control block, a block for controlling the number of generations, a J-input OR element, an AND block. in interval form to the distribution law with a random frequency of occurrence of values within the intervals.

The disadvantage of this device is the formation of a limited sequence of given values of the data set, which is due to the approach implemented in the device to setting the distribution law, which involves specifying the absolute values of the required number of observations of the elements of a given data set at the output of the device for each of the intervals.

A random sequence generator is known according to RF Patent No. 2313125 "Random Sequence Generator" IPC G06F 7/58, H03K 3/84, G07C 15/00 published on 20.12.2007, Bull. No. 35. This device contains a source of random numbers, a random access memory, an N-bit selector-multiplexer, K P-bit registers, K comparison units, a priority encoder, N inverters. The device provides the formation of an unlimited sequence of set values of the data set, the frequency of occurrence of which is determined only by the given distribution law.

However, this device has a drawback - a narrow scope, limited by the possibility of modeling discrete random processes, characterized by the absence of a probabilistic connection between the states of a random process (between the generated values of a given data set), the absence of a probabilistic dependence of each next state of a random process from the previous one.

Of the known closest analogue (prototype), in terms of its technical essence, the claimed device is a random sequence generator according to RF Patent No. No. 6. The prototype device contains a source of random numbers, random access memory, the first and second selectors-multiplexers, K storage blocks of interval boundaries, K comparison blocks, priority encoder, N inverters, first and second registers. The prototype device provides the formation of the values of the elements of a random process, taking into account the probabilistic relationships between the states of this process.

However, the prototype has a drawback - a relatively low reliability of the generated sequence of given values of the data set under conditions typical for random processes occurring in real systems, in real computing, when the states of the generated random process depend not only on the quantitatively given values of the boundaries of the intervals of these states, but and from these values, given taking into account the complex uncertainty, which manifests itself simultaneously and collectively in the form of unreliability and fuzziness of information.

This device allows you to obtain the current probabilistic-temporal values of the elements of a given data set, taking into account the quantitatively specified values of the upper boundaries of the intervals into which the set of addresses of this data set is divided, determined at each subsequent step of the device operation by the probabilities of transitions of a random process from state to state, while as in the study of many random processes occurring in real systems, neuro-fuzzy models [1-6] are widely used, objectively based on parallel, complex data processing, given taking into account the complex uncertainty of the values of the boundaries of the intervals for various states of the random process, where these values are not only quantitative (reliably identifiable), but also simultaneously unreliable and fuzzy, traditionally described with the involvement of the mathematics of neuro-fuzzy networks (NFN), where the neural network component of the NNN is responsible for eliminating the aspect of complex uncertainty of the data unreliability type, and the functions of eliminating the aspect of complex uncertainties of the fuzzy type in the framework of the NNN are implemented within the framework of a component based on algorithms for the transformation of fuzzy sets.

The aim of the invention is to develop a device that improves the reliability of the generated sequence by providing the possibility of generating a given data set, taking into account both the quantitatively specified values of the upper boundaries of the intervals into which the set of addresses of this data set is divided, and taking into account the complex uncertainty of the values of these interval boundaries for various states of a generated random process, the creation of a random sequence generator capable of generating with high reliability the values of the elements of a random process that characterizes the real behavior of a complex computing system, when the upper boundaries of the intervals for various states of this random process are both quantitatively and qualitatively - both unreliable and fuzzy physical meaning.

This goal is achieved by the fact that in a known random sequence generator containing K blocks of comparison, where K≥2 is the maximum possible power of a given set of generated values, the first selector-multiplexer, the second selector-multiplexer, random access memory, priority encoder, N inverters, where N=[log ₂ K] is the number of binary digits sufficient to address the elements of a given set of generated values, K blocks for storing the boundaries of the intervals, made in the form of random access memory devices, the first register, the second register and the source of random numbers, an additional block of neuro- fuzzy network, designed to check and identify the values of the upper boundaries of the intervals (into which the set of addresses of a given data set is divided, determined at each subsequent step by the probabilities of transitions of a random process from state to state), set and identified taking into account the complex uncertainty of these values, as well as for processing and transformation of such initial data, given simultaneously in an unreliable and fuzzy form, to a form that allows you to reliably and unambiguously identify and interpret the values of the initial data on the boundaries of the intervals for various states of the process under study. In this case, the P-bit, where P>N, the output "Random number" of the source of random numbers is connected to the P-bit inputs "Random number" of K comparison blocks, the P-bit output of the k-th storage block of interval boundaries, where k=1, 2, ..., K, connected to the P-bit input "Upper limit" of the k-th comparison block, the output "Comparison result" of the k-th comparison block is connected to the k-th inverse input of the priority encoder, the n-th inverse output of which, where n=1, 2, ..., N, connected to the input of the n-th inverter, the selection input of the first selector-multiplexer is connected to the control input of the random number source, to the selection input of the second selector-multiplexer and is the control input of the generator, the first N-bit information the input of the first selector-multiplexer is the first N-bit address input of the generator, the N-bit output of the first selector-multiplexer is connected to the N-bit address input of the RAM, the inverse inputs "Read/Write" and "Crystal Select" are respectively the first input “Read/Write” and the first input “Crystal Select” of the generator, M-bit output of random access memory, where M≥2 is the number of bits sufficient to represent the maximum value from among the values included in the given set of generated values, is M- bit output "Result" of the generator, the M-bit information input of the RAM is the M-bit information input of the generator. In this case, the P-bit information input of the k-th storage block of the interval boundaries is the k-th P-bit information input of the generator, the n-th bit of the N-bit information input of the second register is connected to the inverse output of the n-th inverter, and the N-bit output of the second register is connected to the second N-bit information input of the first selector-multiplexer and to the second N-bit information input of the second selector-multiplexer, the first N-bit input of which is the second N-bit address input of the generator, and its N-bit output is connected to The N-bit direct input of the neuro-fuzzy network block, the N-bit direct output of which is connected to the N-bit information output of this block and connected to the N-bit information input of the first register, the N-bit neuro-fuzzy input of the neuro-fuzzy network block is An N-bit neuro-fuzzy input of the device, the N-bit output of the first register is connected to the N-bit address inputs of K blocks for storing the boundaries of the intervals, the corresponding inverse inputs "Crystal selection" and "Read / write" of which are combined with each other and are, respectively, the second input “Crystal Select” and the second “Read/Write” input of the generator, the initialization inputs of the first and second registers are respectively the first and second inputs of the “Installation” of the generator, and the reset inputs of the first and second registers are combined and are the “Reset” input of the generator.

A block of a neuro-fuzzy network designed to check and identify the values of the upper boundaries of the intervals (into which the set of addresses of a given data set is divided, determined at each subsequent step by the probabilities of transitions of a random process from state to state), specified and identified taking into account the complex uncertainty of these values, as well as for processing and transforming such initial data, given simultaneously in an unreliable and fuzzy form, to a form that allows you to reliably and unambiguously identify and interpret the values of the initial data on the boundaries of the intervals for various states of the process under study, consists of a counter, a storage register, a neuro-fuzzy programmable a calculator and a storage element, the N-bit output of which is the N-bit information output of the neuro-fuzzy network block, the N-bit input of the neuro-fuzzy programmable calculator is connected to the N-bit information output of the storage register and is the N-bit neuro-fuzzy input of the block neuro-fuzzy network and an N-bit neuro-fuzzy input of the device, the enable input of outputs A of the neuro-fuzzy programmable computer is connected to the enable output of the storage register, N outputs A of the neuro-fuzzy programmable computer are connected to the corresponding N inputs of the storage element, N-bit direct the output of the storage register is the N-bit direct output of the neuro-fuzzy network block, the N-bit input of the storage register is connected to the N-bit output of the counter, the N-bit input of which is the N-bit direct input of the neuro-fuzzy network block.

Thanks to a new set of essential features, due to the introduction of a neuro-fuzzy network block, which provides preliminary identification and selection of the values of the upper limits of the intervals identified taking into account the complex uncertainty of these values and making a decision about their mathematical nature, as well as recording and storing the results of the analysis of the values of the upper limits intervals specified simultaneously in an unreliable and fuzzy form, and mathematically correct computational transformation and recognition of these values using neuro-fuzzy mathematical methods, in the claimed random sequence generator, it is possible to increase the reliability of the generated sequence of given data set values by providing the possibility of their generation taking into account the presence of not only a quantitative measure of identification of the states of a random process, but also a qualitative measure - complex uncertainty, i.e., both unreliably and fuzzy set values of the upper boundaries of the state intervals for a specific step of modeling this discrete random process.

The claimed device is illustrated by drawings, in which:

in fig. 1 - block diagram of a random sequence generator;

in fig. 2 - block diagram of the neuro-fuzzy network 11;

in fig. 3 - scheme for constructing a neuro-fuzzy network;

in fig. 4 is a block diagram of the comparison block.

The random sequence generator (see Fig. 1) consists of the first selector-multiplexer 1, the second selector-multiplexer 2, the first register 3, the source of random numbers 4, random access memory 5, K storage blocks of interval boundaries 6 ₁ -6 _K , K comparison blocks 7 ₁ -7 _K , priority encoder 8, N inverters 9 ₁ -9 _N , second register 10 and neuro-fuzzy network block 11.

) и «Выбор кристалла» 51 (вход

) которого являются соответственно первым входом «Чтение/запись» 02 и первым входом «Выбор кристалла» 04 генератора. Причем М-разрядный выход 55 (выход для разрядов C₁-C_M) оперативного запоминающего устройства 5 является М-разрядным выходом «Результат» 013 генератора, М-разрядный информационный вход 53 (вход для разрядов D₁-D_M) оперативного запоминающего устройства 5 является М-разрядным информационным входом 03 генератора. При этом Р-разрядный информационный вход 63_k (вход для разрядов D₁-D_P) k-го блока хранения границ интервалов 6_k является k-ым Р-разрядным информационным входом 09_k генератора, n-ый разряд N-разрядного информационного входа 101 второго регистра 10 соединен с инверсным выходом 92_n n-го инвертора 9_n, а N-разрядный выход 104 второго регистра 10 подключен ко второму N-разрядному информационному входу 120 (входу для разрядов B₁-B_N) первого селектора-мультиплексора 1 и ко второму N-разрядному информационному входу 22 (входу для разрядов B₁-B_N) второго селектора-мультиплексора 2, первый N-разрядный вход 21 которого является вторым N-разрядным адресным входом 05 генератора, а его N-разрядный выход 24 соединен с N-разрядным прямым входом 111 блока нейро-нечеткой сети 11, N-разрядный прямой выход 112 которого соединен с N-разрядным информационным выходом 113 блока нейро-нечеткой сети 11 и подключен к N-разрядному информационному входу 31 (входу для разрядов D₁-D_N) первого регистра 3, N-разрядный нейро-нечеткий вход 114 блока нейро-нечеткой сети 11 является N-разрядным нейро-нечетким входом 014 устройства, N-разрядный выход 34 первого регистра 3 соединен с N-разрядными адресными входами 62₁-62_K K блоков хранения границ интервалов 6₁-6_K, соответствующие инверсные входы «Выбор кристалла» 61₁-61_K (входы

) и «Чтение/запись» 64₁-64_K (входы

) которых объединены между собой и являются соответственно вторым входом «Выбор кристалла» 08 и вторым входом «Чтение/запись» 010 генератора. Входы инициализации 33 и 103 (входы С) первого 3 и второго 10 регистров являются соответственно первым 06 и вторым 012 входом «Установка» генератора, а входы сброса 32 и 102 (входы R) первого 3 и второго 10 регистров объединены и являются входом «Обнуление» 011 генератора.The elements are interconnected as follows (see Fig. 1). R-bit output "Random number" 42 of the source of random numbers 4 is connected to P-bit inputs "Random number" 71 ₁ -71 _K Compare units 7 ₁ -7 _K , P-bit output 65 _{k of} the k-th storage block of interval boundaries 6 _k , where k=1, 2, ..., K, is connected to the P-bit input "Upper Limit" 72 _k of the k-th comparison unit 7 _k . The output "Comparison result" 73 _k of the k-th comparison block 7 _k is connected to the k-th inverse input 81 _k of the priority encoder 8, the n-th inverse output 82 _n of which, where n=1, 2, ..., N, is connected to the input 91 _n of the nth inverter 9 _n . The selection input 130 (SE input) of the first selector-multiplexer 1 is connected to the control input 41 of the random number source 4 with the selection input 23 (SE input) of the second selector-multiplexer 2 and is the control input 07 of the generator, the first N-bit information input 110 (input for bits A ₁ -A _N ) of the first selector-multiplexer 1 is the first N-bit address input 01 of the generator, the N-bit output 140 (output for bits Q ₁ -Q _N ) of the first selector-multiplexer 1 is connected to the N-bit address input 52 (input for bits A ₁ -A _N ) random access memory 5, inverse inputs "Read / write" 54 (input

) and "Crystal Select" 51 (input

) of which are respectively the first input "Read/Write" 02 and the first input "Crystal Select" 04 of the generator. Moreover, the M-bit output 55 (output for bits C ₁ -C _M ) of the random access memory device 5 is the M-bit output "Result" 013 of the generator, the M-bit information input 53 (input for the bits D ₁ -D _M ) of the random access memory device 5 is the M-bit information input 03 of the generator. In this case, the P-bit information input 63 _k (input for bits D ₁ -D _P ) of the k-th block of storage of interval boundaries 6 _k is the k-th P-bit information input 09 _k of the generator, the n-th bit of the N-bit information input 101 of the second register 10 is connected to the inverse output 92 _n of the n-th inverter 9 _n , and the N-bit output 104 of the second register 10 is connected to the second N-bit information input 120 (input for bits B ₁ -B _N ) of the first selector-multiplexer 1 and to the second N-bit information input 22 (input for bits B ₁ -B _N ) of the second selector-multiplexer 2, the first N-bit input 21 of which is the second N-bit address input 05 of the generator, and its N-bit output 24 is connected with an N-bit direct input 111 of the block of the neuro-fuzzy network 11, the N-bit direct output 112 of which is connected to the N-bit information output 113 of the block of the neuro-fuzzy network 11 and is connected to the N-bit information input 31 (input for bits D ₁ -D _N ) of the first register 3, the N-bit neuro-fuzzy input 114 of the block of the neuro-fuzzy network 11 is the N-bit neuro-fuzzy input 014 of the device, the N-bit output 34 of the first register 3 is connected to the N-bit address inputs 62 ₁ -62 _K K interval boundaries storage blocks 6 ₁ -6 _K , corresponding inverse inputs "Crystal selection" 61 ₁ -61 _K (inputs

) and "Read/Write" 64 ₁ -64 _K (inputs

) which are combined with each other and are respectively the second input "Crystal Select" 08 and the second input "Read / Write" 010 of the generator. Initialization inputs 33 and 103 (inputs C) of the first 3 and second 10 registers are, respectively, the first 06 and second 012 input "Installation" of the generator, and the reset inputs 32 and 102 (inputs R) of the first 3 and second 10 registers are combined and are the input "Reset » 011 generator.

Complex uncertainty is a type of description of semi-structured and difficult to formalize data, in which the lack of our knowledge about the possible state of the environment and the object of research, as well as the fact that we completely or partially do not have information about the values of the interval boundaries for various states of the generated random process, simultaneously and cumulatively manifest themselves as unreliability and fuzziness of the initial data on the states of the process under study [7-10].

The physical and mathematical essence of the values of the boundaries of the intervals is the numerical limits of the boundaries of the states of a random process. The values of the boundaries of the intervals are a set of binary codes that determine the threshold, boundary values of the probability of occurrence of the corresponding elements of a given data set (data set in a given interval). To set the intervals, their upper limits, the values of the upper limits of the intervals are indicated.

The number "N, (N=[log ₂ K]; N≥1)" (blocks, bits, inputs, outputs, etc.) is determined in accordance with the number of bits sufficient to address the elements of a given set of generated random sequence values (i.e., for addressing elements of a data set), with the number of calculators (neurons) in each layer of the NNN and, as a rule, ranges from 2 (two) to 10 (ten).

The number "K, (K≥2)" characterizes the maximum possible power of a given set of generated values of a random sequence (i.e., the number of elements in a given data set) and, as a rule, ranges from 2 (two) to 500 (five hundred).

The number "M, (M≥2)" characterizes the number of binary digits sufficient to represent the maximum value from the number of values that are part of a given set of generated values of a random sequence (i.e., to represent the values of elements of a given data set) and, as a rule, , ranges from 2 (two) to 10 (ten).

The number "P, (P>N)" characterizes the capacity of the information inputs of the generator and is determined in accordance with the number greater than the number of binary digits sufficient to address the elements of a given set of generated random sequence values and, as a rule, ranges from 3 (three) up to 11 (eleven).

The neuro-fuzzy network block 11 shown in FIG. 2 is designed to check and identify the values of the upper boundaries of the intervals (into which the set of addresses of a given data set is divided, determined at each subsequent step by the probabilities of transitions of a random process from state to state), set and identified taking into account the complex uncertainty of these values, as well as for processing and transformation of such initial data, given simultaneously in an unreliable and fuzzy form, to a form that allows one to reliably and unambiguously identify and interpret the values of the initial data on the boundaries of the intervals for various states of the process under study.

The neuro-fuzzy network block 11 (Fig. 2) consists of a counter 11.1, a storage register 11.2, a neuro-fuzzy programmable calculator 11.3 and a memory element 11.4, the N-bit output 11.4-2 of which is the N-bit information output 113 of the neuro-fuzzy block network 11, N-bit input 11.3-1 (I ₁ -I _N ) neuro-fuzzy programmable calculator 11.3 is connected to the N-bit information output 11.2-1 of the storage register 11.2 and is an N-bit neuro-fuzzy input 114 block neuro-fuzzy network 11 and N-bit neuro-fuzzy input 014 of the device, input 11.3-2 (OE _I ) enable outputs And neuro-fuzzy programmable computer 11.3 is connected to the enable output 11.2-3 of the storage register 11.2, N outputs A (A ₁ -A _N ) of the neuro-fuzzy programmable computer 11.3 are connected to the corresponding N inputs 11.4-1 ₁ -11.4-1 _N of the storage element 11.4, the N-bit direct output 11.2-4 of the storage register 11.2 is the N-bit direct output 112 of the block of the neuro-fuzzy network 11, The N-bit input 11.2-2 of the storage register 11.2 is connected to the N-bit output 11.1-2 of the counter 11.1, the N-bit input 11.1-1 of which is the N-bit direct input 111 of the neuro-fuzzy network block 11.

The counter 11.1 of the block of the neuro-fuzzy network 11 is designed to register and sequentially compare, by the number of digits received in the binary data code, characterizing the values of the upper limits of the intervals, and convert this data from the parallel code to the serial one. The technical implementation of the counter 11.1 is possible on the basis of a commercially available programmable summing counter with the function of random access memory (RAM) and controlled reset, the scheme of which is known and described, for example, in [Ugryumov E.P. Digital circuitry: textbook. allowance for universities. - 3rd ed., revised. and additional - St. Petersburg: BHV-Petersburg, 2010. - 816 p., S. 246-247, fig. 3.46].

The storage register 11.2 of the block of the neuro-fuzzy network 11 is designed to make a decision about the mathematical nature of the initial data (threshold, boundary values of the probability of occurrence of the generated data set in a given interval), given in the form of a binary code - the values of the upper limits of the intervals are set quantitatively or taking into account complex uncertainty these values, i.e., are given simultaneously in an unreliable and fuzzy form and require additional verification and confirmation. The storage register 11.2 can be technically implemented on the basis of a commercially available standard storage register on flip-flops described in the literature [Fundamentals of Electronics: a textbook for open source software / O.V. Milovzorov, I.G. Pankov. - 5th ed., revised. and additional - M.: Yurayt Publishing House, 2016. - 407 p. pp. 232-233, fig. 4.3.8].

The neuro-fuzzy programmable calculator 11.3 of the block of the neuro-fuzzy network 11 is designed to carry out the procedure of neuro-fuzzy transformation of the initial data - the values of the upper limits of the intervals, set simultaneously in an unreliable and fuzzy form, to a form that allows you to reliably identify and interpret the values of these upper limits of the intervals for a specific step of modeling discrete random processes. The neuro-fuzzy programmable computer 11.3 of the block of the neuro-fuzzy network 11 from the point of view of the theory of NNS is a hardware-software analogue of a five-layer neuro-fuzzy network of direct signal propagation, such as ANFIS (Adaptive-Network-Based Fuzzy Inference System - adaptive fuzzy inference network ) [12]. A variant of such an NNS is shown in Fig. 3. From the point of view of the theory of NNS, the neuro-fuzzy programmable calculator 11.3 of the unit of the neuro-fuzzy network 11 performs the following functions: fuzzification, i.e., transformation of numerical input values in the degree of correspondence with linguistic variables (functions of the first layer of the NNS); recording and storing fuzzy rules, which are a set of fuzzy rules of the "If-Then" type (functions of the second layer of the NNN); recording and storing data that determine the membership functions of fuzzy sets used in fuzzy rules (functions of the third layer of the NNN); decision-making, i.e., performing an inference operation based on the existing rules (functions of the fourth layer of the NNN) and defuzzification functions aimed at converting the results of the inference into numerical values of the initial data on the boundaries of the intervals for various states of the process under study (functions of the fifth, output layer of the NNN ). Neuro-fuzzy programmable computer 11.3 can be technically implemented on the basis of commercially available microprocessor section (MPS or MPS - Micro-Processoring Section) type MPS K1804BC1, described, for example, in [Ugryumov E.P. Digital circuitry: textbook. allowance for universities. - 3rd ed., revised. and additional - St. Petersburg: BHV-Petersburg, 2010. - 816 p., S. 371-376, fig. 6.14].

The storage element 11.4 of the block of the neuro-fuzzy network 11 is designed to record and store the verified and confirmed results of the analysis of the values of the upper bounds of the intervals, as well as to convert these data from serial to parallel code. The memory element 11.4 can be technically implemented on the basis of a commercially available programmable dynamic random access memory with N inputs and N-bit output, in accordance with the description presented in [Fundamentals of Electronics: textbook for SPO / O.V. Milovzorov, I.G. Pankov. - 5th ed., revised. and additional - M.: Yurayt Publishing House, 2016. - 407 p. pp. 229-231, fig. 4.3.2].

The first selector-multiplexer 1, which is included in the general block diagram of the random sequence generator, is designed to switch to its N-bit output signals of one of two groups: from the first N-bit information input 110 (input for bits A ₁ -A _N ) or from second N-bit information input 120 (input for bits B ₁ -B _N ). The construction scheme of the first selector-multiplexer 1 is known, it can be technically implemented on the basis of a commercially available N-bit multiplexer, as shown in [Ugryumov E.P. Digital circuitry: textbook. allowance for universities. - 3rd ed., Rev. and additional - St. Petersburg: BHV-Petersburg, 2010. - 816 p., S. 94-95].

The second selector-multiplexer 2, which is included in the general block diagram of the random sequence generator, is designed to switch to its N-bit output signals of one of two groups: from the first N-bit input 21 (input for bits A ₁ -A _N ) or from the second N-bit information input 22 (input for bits B ₁ -B _N ). It is responsible for recording the values of the upper boundaries of the intervals, the value of which can be set quantitatively or both unreliably and fuzzy, the boundaries into which the set of addresses of a given data set is divided, determined at each subsequent step by the given probabilities of transitions of the random process from state to state. The construction scheme of the second selector-multiplexer 2 is known and similar to the scheme of the first selector-multiplexer 1. It can be technically implemented on the basis of a commercially available N-bit multiplexer, as shown in [Ugryumov E.P. Digital circuitry: textbook. allowance for universities. - 3rd ed., revised. and additional - St. Petersburg: BHV-Petersburg, 2010. - 816 p., S. 94-95].

The first register 3, included in the general block diagram of the random sequence generator, is designed to register and store binary codes that determine the generation of a sequence of values of a given data set, as well as to register and store new values for each subsequent step, verified and confirmed (with the help of neuro- fuzzy network) upper bounds of intervals, the value of which can be specified quantitatively or both unreliably and fuzzy, and dynamically changes according to the methods of Markov chains, depends on the probability of transition of a random process from state to state and corresponds to the values of the probabilities of observing the corresponding elements of a given set required at this step data. The description of the work and the scheme of such registers are known and are given, for example, in the book [Averchenkov O.E. Circuit engineering: hardware and software. - M.: DMK Press, 2012. - 588 p., S. 443-445].

The source of random numbers 4, which is included in the general block diagram of the random sequence generator, is designed to generate P-bit random numbers, is known, and can be technically implemented, as shown, for example, in [Ugryumov E.P. Digital circuitry: textbook. allowance for universities. - 3rd ed., revised. and additional - St. Petersburg: BHV-Petersburg, 2010. - 816 p., S. 260-261].

Random access memory 5, which is part of the general block diagram of the random sequence generator, is designed to store the values of the elements of a given data set. The scheme for constructing random access memory is known, it can be technically implemented on the basis of commercially available random access memory, as shown in [Averchenkov O.E. Circuit engineering: hardware and software. - M: DMK Press, 2012. - 588 p., S. 246-247].

Blocks for storing the boundaries of intervals 6 ₁ -6 _K , included in the general block diagram of the random sequence generator, are designed to store binary codes that define new, verified and confirmed (using a neuro-fuzzy network) upper boundaries of the intervals, i.e., boundary values the probability of occurrence of the corresponding elements of a given data set (data set in a given interval). They are made in the form of random access memory devices, the structure of their construction is known, they can be technically implemented on the basis of commercially available random access memory devices, as shown in [Averchenkov O.E. Circuit engineering: hardware and software. - M.: DMK Press, 2012. - 588 p., S. 246-247].

блока сравнения 7к известны, описаны в прототипе (см. патент РФ №2542903 «Генератор случайной последовательности» МПК G06F 7/58, G06F 1/02, опубликованный 27.02.2015, Бюл. №6, фиг. 4), структурная схема проиллюстрирована, на примере k-го блока сравнения 7_k, на фиг. 4.Comparison units 7 ₁ -7 _K included in the overall block diagram of the random sequence generator are designed to compare the random value generated by the source of random numbers, and the value at the output of the corresponding register, as well as to generate the result of the comparison. The principle of operation and structure of any k-th

comparison block 7k are known, described in the prototype (see RF patent No. 2542903 "Random sequence generator" IPC G06F 7/58, G06F 1/02, published on February 27, 2015, Bull. No. 6, Fig. 4), the block diagram is illustrated, on the example of the k-th comparison block 7 _k , in FIG. 4.

The priority encoder 8, which is part of the general circuit of the random sequence generator, is designed to convert the value of a logical zero at one of its inputs into the corresponding binary code at its output, and the conversion is carried out taking into account the priorities determined by the input number. The implementation scheme of the priority encoder is known and described in detail, for example, in [Averchenkov O.E. Circuit engineering: hardware and software. - M.: DMK Press, 2012. - 588 p., S. 414-418].

Inverters 9 ₁ -9 _N included in the general scheme of the random sequence generator are designed to invert signals from the inverse outputs of the priority encoder. The inverter implementation scheme is known and described in detail, for example, in [Averchenkov O.E. Circuit engineering: hardware and software. - M.: DMK Press, 2012. - 588 p., S. 426].

The second register 10, included in the general circuit of the random sequence generator, is designed to store binary codes that determine the probabilities of occurrence of the corresponding elements of a given data set. It is responsible for registering and storing the values of the previous (for the last step) upper limits of the intervals checked and confirmed using the NNN (obtained taking into account the presence of not only quantitatively, but also at the same time unreliably and fuzzy given values of these limits), the value of which dynamically changes according to the methods of Markov chains, depends on the probability of transition of a random process from state to state and corresponds to the values of the probabilities of observing the corresponding elements of a given data set required at the last step. It is similar to the first register 3, the description of the work and the scheme of such registers are known and are given, for example, in the book [Averchenkov O.E. Circuit engineering: hardware and software. - M.: DMK Press, 2012. - 588 p., S. 443-445].

In the claimed device, the generation of set values of a data set, taking into account both the quantitative values of the boundaries between the generated values of this data set (boundaries of the intervals), and taking into account the complex uncertainty of these values, is carried out in stages. The first step is the generation of the given values of the data set, taking into account the quantitatively specified upper bounds of the intervals, implemented on the basis of the distribution law, which is specified by specifying the required probability of the occurrence of the corresponding element of the given data set. In this case, the approach [11] is applied, based on the use of a source of random numbers distributed uniformly in the range [0; 1). This range is divided into a set of intervals, the number of which corresponds to the number of elements in the given data set. The values of the intervals correspond to the values of the required probabilities of observing the corresponding elements of a given data set. The probability of hitting a random number generated by a random number source inside each interval is equal to its length. The interval number is used as an address for retrieving an element of a given data set from random access memory, and for setting intervals, their upper bounds are indicated.

The initial assignment of the initial probabilities of the occurrence of elements of a given data set, at the first step, without taking into account the complex uncertainty of the values of the upper boundaries of the intervals, is carried out as follows.

K - количество элементов заданного набора данных, а В={b₁, b₂, …, b_k, …, b_K} - множество количественно заданных соответствующих значений начальных верхних границ интервалов. Тогда начальное, количественно, полно, достоверно и однозначно заданное значение верхней границы k-го интервала будет равно сумме значений вероятностей появления элементов из заданного набора данных от 1-го до k-го:Let Н={h ₁ , h ₂ , …, h _k , …, h _K } be the set of required probabilities of occurrence of elements of a given data set, where

K is the number of elements of the given data set, and B={b ₁ , b ₂ , …, b _k , …, b _K } is the set of quantitatively specified corresponding values of the initial upper bounds of the intervals. Then the initial, quantitatively, completely, reliably and uniquely specified value of the upper boundary of the kth interval will be equal to the sum of the probabilities of the occurrence of elements from the given data set from the 1st to the kth:

Thus, the range [0; 1) at the first generation step will be divided into the following intervals:

A P-bit source of random numbers with a uniform distribution law is used. In this case, the P-bit binary code generated at the output of the random number source, and the P-bit binary code used to set the initial (for the first step) upper bound of the interval, are considered quantitatively as numbers from the range [0; 1) without specifying the whole part.

At the same time, either within the framework of generating random numbers with a uniform distribution law, or from the outside, from the device operator, the values of the upper limits of the intervals can be set, which have a complex uncertainty, i.e., set qualitatively, and having both an unreliable and fuzzy physical meaning .

If a set of required probabilities of occurrence of elements of a given data set is given H={0.125; 0.125; 0.25; 0.5}, then the quantitatively specified values of the upper limits of the intervals will be equal to 0.125, 0.25, 0.5, and 1, respectively. numbers 0.0010, 0.0100, 0.1000, 0.1111. And only the fractional part of these numbers is entered into the device: the values 0010, 0100,1000 and 1111, respectively.

Similarly, if the random number source generated the value 1101 (in decimal representation 0.8125), then representing this value as a fractional part of a number from the range [0; 1), i.e. as 0.1101, and given the set of quantitatively specified (for the first stage) initial upper bounds of the intervals from the previous paragraph, it can be seen that the random value fell into the interval limited by the numbers 0.1000 and 0.1111.

In order for the random sequence generator to be able to simulate discrete random processes occurring in real systems, taking into account the values of the boundaries of the intervals for various states of the random process, set both quantitatively and qualitatively - taking into account the complex uncertainty of the values of these boundaries between the generated values of a given set data, in accordance with the theory of Markov processes [13], first set the initial values of the probability of transitions of a random process from state to state.

At the same time, at each subsequent step of the generator operation, the probabilities of occurrence of elements of a given data set will have new values of the upper bounds of the intervals, determined in accordance with the Markov model, and the generated random values at each new step will fall into a new one (only for this step) interval. This step-by-step refinement, identification of the values of the set of upper bounds of the intervals makes it possible to obtain the values of an element from a given data set, taking into account the presence of quantitative values of these upper bounds of the intervals for a specific step of modeling discrete random processes.

In addition to quantitative threshold values, accounting for the complex uncertainty of the values of the upper boundaries of the intervals for a particular step, implemented in the proposed random sequence generator, is carried out within the framework of procedures based on the well-known results of research in the field of the theory of neuro-fuzzy networks, described in [1-6]. The analysis of works [1-6] makes it possible to form a mathematically correct neuro-fuzzy algorithm for reducing the initial data given simultaneously in an unreliable and fuzzy form - the values of the upper boundaries of the intervals for various states of the Markov chain, to the nearest complete and unambiguous given set.

Thus, within the framework of modeling discrete random processes, which are formed taking into account both quantitatively and qualitatively - both unreliably and fuzzy identified values of the upper boundaries of the intervals for various states of the Markov chain, a number of characteristics of the generated process are modeled on the basis of parametrically specified initial data, by traditional methods, and modeling taking into account the complex uncertainty of the values of these boundaries of the intervals (i.e., taking into account both unreliably and fuzzy threshold values of states for a specific step of modeling discrete random processes) - by successive transformations using the methods of the theory of NNN, is reduced to the possibility of their relatively parametric modeling, i.e. the transition from the simultaneously unreliable and fuzzy task of identifying the upper boundaries of the intervals to the parametric one is carried out.

It is known [1-6] that from the point of view of additional verification and confirmation using neuro-fuzzy verification procedures for the values of the upper boundaries of the intervals for a specific step, taking into account the complex uncertainty of these initial data for various states of the Markov chain, it is possible to determine these values given as quantitatively and qualitatively - both unreliable and fuzzy. This possibility is implemented using neuro-fuzzy computational methods and algorithms that allow, through sequential mathematical neuro-fuzzy transformations, to make the transition from unreliably and fuzzy upper threshold values of states to a form suitable for unambiguously making a reliable decision about the values of these upper limits of intervals for a particular step of modeling discrete random processes.

At the same time, the values of the upper limits of the intervals for various states of the Markov chain and for a specific step in the modeling of discrete random processes can be additionally checked and confirmed on the basis of mathematical decision-making methods in semi-structured problems - neuro-fuzzy computational methods and algorithms that can be implemented quite simply in hardware. in the form of NNS. Neuro-fuzzy networks, computational methods and algorithms based on them will combine neural networks and fuzzy logic, collect the best properties of both methods, and at the same time get rid of their problems. On the one hand, such structures include the computational power and learning ability of neural networks, and on the other hand, the intellectual capabilities of neural networks are enhanced by fuzzy decision-making rules inherent in the classical ways of “thinking” [1, 2, 4].

A computational neuro-fuzzy algorithm of this class consists of five layers of N calculators (neurons) in each. At the same time, the number N of calculators (neurons) of the input, intermediate and output layers corresponds to the addressing code of the elements of the data set, is equal to N=[log ₂ K] and depends on the number of binary digits sufficient to describe the values of the upper bounds of the intervals for various states of the Markov chain, from the number of experts and the corresponding number of inputs of devices for the hardware implementation of the neuro-fuzzy network. Moreover, N calculators (neurons) in each layer of the NNN (in each branch of the neuro-fuzzy algorithm) are responsible for converting simultaneously unreliably and fuzzy recognized and certain values of the upper threshold values of the states to a form that allows you to unambiguously and reliably identify and interpret the values of these upper limits of the intervals for specific step of modeling discrete random processes.

In our case, N can take values from 2 (two) to 10 (ten), corresponding to the number of inputs of the MPS K1804BC1 microprocessor section, on the basis of which the programmable computer 11.2 of the neuro-fuzzy network block 11 is built.

Neuro-fuzzy networks allow for additional verification and confirmation of the validity of the characteristics of the simulated generated process, and the derivation of these characteristics - the values of the upper boundaries of the intervals, is carried out on the basis of the fuzzy logic apparatus, and the parameters of membership functions are adjusted using neural network training algorithms. In this case, the fuzzy control module is presented in the form of a multilayer network, in which the layers act as elements of the fuzzy inference algorithm [1, 2, 4].

The classic fuzzy inference algorithm in the NNN consists of five functional stages: the fuzzification stage, which transforms numerical input values into the degree of correspondence to linguistic variables; the stage of obtaining, recording and storing fuzzy rules, containing a set of fuzzy rules of the "If-Then" type; the stage of obtaining, recording and storing data, which determines the membership functions of fuzzy sets used in fuzzy rules; the decision-making stage, within which inference operations are carried out based on the existing rules, as well as the defuzzification stage, at which the inference results are converted into numerical values of the upper boundaries of the intervals [1, 2].

The input image is received at the input of the neuro-fuzzy network - data characterizing the numerical input values of the upper boundaries of the intervals and recognized, determined both quantitatively and qualitatively - both unreliably and fuzzy. It is determined which of the data characterizing the values of the upper limits of the intervals are quantitatively recognized at a given time, and which data are identified both unreliably and indistinctly. For the purpose of additional verification and confirmation, using the procedures of neuro-fuzzy verification, the values of the upper limits of the intervals, it is necessary to mathematically correct, using the NNN, to convert the data recognized both unreliably and fuzzy.

для этих сигналов. Параметры функций принадлежности для конкретных значений верхних границ интервалов для каждого конкретного шага становятся весами связей для нейронов первого слоя ННС, и они могут модифицироваться в процессе обучения. В качестве функций принадлежностей входных и выходных переменных - значений верхних границ интервалов, используется функция Гаусса в видеIn this case, the first (input) layer of the NNN implements the first stage, the fuzzification procedure - the transformation of numerical input values in the degree of compliance with linguistic variables, characterized by membership functions for each input variable - the values of the upper bounds of the intervals for each specific step. The first input of the first layer receives input signals that characterize a specific value at the same time unreliably and fuzzy set value of the upper limit of the intervals x ₁ , the second - the opinions of experts x ₂ about this value. At the output of the layer, we obtain the value of the membership function

for these signals. The parameters of the membership functions for specific values of the upper limits of the intervals for each specific step become the weights of connections for the neurons of the first layer of the NNN, and they can be modified in the learning process. As the membership functions of the input and output variables - the values of the upper boundaries of the intervals, the Gaussian function is used in the form

where q _i , w _j are the parameters of the membership function for specific values of the upper bounds of the intervals that require adjustment in the learning process of the NNN, x _n are the values of the upper bounds of the intervals coming to the input of the NNN.

The configuration of the links of the second layer corresponds to the structure of the rules:

и х₂ есть W₁,Rule R ₁ : if x ₁ exists

and x ₂ is W ₁ ,

и х₂ есть W₂,Rule R ₂ : if x ₁ exists

and x ₂ is W ₂ ,

и х₂ есть W_n,R _n rule: if x ₁ is

and x ₂ is W _n ,

- нечеткие множества, элементы которых описывают одновременно не-Where

- fuzzy sets, the elements of which simultaneously describe indistinct

Reliably and indistinctly specified values of the upper limits of the intervals.

Then the rule (R ₁ , R ₂ , …, R _n ) can be represented as a fuzzy implication (combining two statements into one).

The second layer implements the inference stage. The number of neurons N in the second layer of the NNN is equal to the number of rules. Each node of the layer is connected to the previous layer in such a way that the node of the second layer of the NNN corresponding to the kth rule is connected to all neurons of the first layer of the NNN corresponding to the fuzzy sets of conditions of this rule. The output value of the second layer will be the weight of the rule:

- нечеткие множества, элементы которых описывают функции принадлежности для конкретных одновременно недостоверно и нечетко заданных значений верхних границ интервалов.Where

- fuzzy sets, the elements of which describe the membership functions for specific, at the same time, uncertainly and fuzzily given values of the upper boundaries of the intervals.

The elements of the third layer carry out the normalization of the degree of execution of the rules and calculate the normalized values of specific upper bounds of the intervals

The clear values of specific upper bounds of the intervals that define the conclusion of each rule are considered in the fourth layer as a fuzzy set with a Gaussian membership function. The adaptive nodes of the fourth layer calculate the contribution of each fuzzy rule to the NNN output by the formula

The fifth layer is the implementation of the defuzzification functions. At the output of the fifth layer, final clear values of the upper boundaries of the intervals are formed.

The stages of functioning of the computational neuro-fuzzy algorithm implemented by the NNN are described in detail, algorithmically and analytically in [1, 2]. At the output of the computational neuro-fuzzy algorithm, we have an output image - data that reliably and unambiguously characterize the belonging of the values of the upper boundaries of the intervals to the space of consistent, verified and confirmed values.

Considered in [4-6] and described in detail in [1, 2], the computational neuro-fuzzy algorithm makes it possible to eliminate the complex uncertainty of the data characterizing the values of the upper limits of the intervals for a particular step, makes it possible to unambiguously recognize, verify and confirm the true values of these upper limits of the intervals, and ultimately, to increase the reliability of the formation of the value of the elements of a random process that characterizes the real behavior of a complex computing system - when the upper boundaries of the intervals for various states of this random process have both quantitatively and qualitatively - both unreliably and fuzzy physical meaning.

In other words, the analysis of the computational neuro-fuzzy algorithm considered in [4–6] and described in detail in [1, 2] allows us to conclude that it is technically possible to implement reliable identification of the essential parameters of the generated data stream, taking into account both quantitatively specified values of the upper boundaries of the intervals, on which the set of addresses of this data stream is divided, and taking into account the complex uncertainty of the values of these interval boundaries for various states of the generated random process. This allows to reduce the uncertainty, and, consequently, to increase the reliability of the generated sequence of set values of the data set.

With this in mind, in the claimed device, the set values of the data set are generated, taking into account the presence of not only a quantitative measure for identifying the states of a random process, but also a qualitative measure - complex uncertainty, i.e., both unreliably and fuzzy set values of the upper boundaries of the state intervals for a specific modeling step this discrete random process.

The technical implementation of the principle of increasing the reliability of recognition of the values of the upper limits of the intervals in the claimed device is carried out by introducing a preliminary analysis and identifying these values of the upper limits of the intervals, set and identified taking into account the complex uncertainty of these values, as well as additional verification and confirmation of these values, set simultaneously in unreliable and fuzzy form, using neuro-fuzzy methods to a form that allows you to reliably and unambiguously identify and interpret the values of these initial data on the boundaries of the intervals for various states of the process under study (in the claimed device, they are implemented within the framework of the neuro-fuzzy network block 11).

The random sequence generator works as follows.

In the initial position, the control input 07, the first 06 and the second 012 inputs "Installation" are set to logical zero values, and the first 02 and second 010 inputs "Read / Write" and the first 04 and second 08 inputs "Crystal selection" - the values of the logical units. The generator operates in the following modes: preparation for generation mode; generation mode.

The mode of operation of the generator, taking into account not only the quantitatively specified values of the upper limits of the state intervals, but also these values, set taking into account their complex uncertainty, is determined by a combination of signals at the control input 07, the first 06 and the second 012 inputs "Installation", the first 02 and the second 010 inputs "Read / write", as well as on the first 04 and second 08 inputs "Crystal selection". To transfer the device to the preparation mode, it is necessary to apply logical zero values to the control input 07 of the generator, the first 04 and second 08 inputs "Crystal selection", the first 02 and second 010 inputs "Read / Write", and to the first 06 and second 012 inputs "Installation » - the value of the logical unit.

In the generation mode, the generator control input 07, the first 02 and second 010 inputs "Read/Write" are set to the values of a logical unit, and the first 04 and second 08 inputs "Crystal selection" and the first 06 and second 012 inputs "Installation" are set to the values of a logical zero.

The following steps are performed in the generator standby mode.

The first step is to enter the set A={ a ₁ , a ₂ , …, a _K } of the values of the elements of a given data set into the random access memory, while:

;0≤ a _k ≤2 ^M -1 - element of the given data set, where

;

M≥2 - the number of binary digits sufficient to represent the values of the elements of a given data set;

K≥2 - the number of elements in a given data set;

N≥1 - the number of binary digits sufficient to address the elements of the data set.

The second step is setting the set B={b ₁ , b ₂ , …, b _K } of quantitative values of the initial upper bounds of the intervals into which the entire set of addresses of the given data set A is divided.

The third step is the initial setting of the set quantitatively С={c ₁ , c ₂ , …, c _K } or, through the N-bit neuro-fuzzy input 114 of the block of the neuro-fuzzy network 11, which is the N-bit neuro-fuzzy input 014 of the device - simultaneously in uncertainly and fuzzy values of new (for a given step) upper boundaries of the intervals into which the set of addresses of the data set A is divided, determined at each subsequent step by the probabilities of transitions of the random process from state to state [13].

и 54

значений логического нуля по первым входам «Выбор кристалла» 04 и «Чтение/запись» 02 соответственно. Затем, по первому N-разрядному адресному входу 01 генератора на первую группу информационных входов 110 (группу входов A₁-A_N) первого селектора мультиплексора 1 подается N-разрядный адрес, по которому должно быть записано значение второго элемента а ₂, а по М-разрядному информационному входу 03 генератора на информационные входы 53 (входы D₁-D_M) оперативного запоминающего устройства 5 подают значение элемента а ₂ и, путем установки значений логического нуля на первых входах «Выбор кристалла» 04 и «Чтение/запись» 02, записывают значение а ₁ в оперативное запоминающее устройство 5. Аналогичным образом в оперативное запоминающее устройство 5 заносятся все K значений элементов заданного набора данных. После чего на первом входе «Чтение/запись» 02 генератора устанавливают значение логической единицы.The first step in preparing the generator for operation includes the following steps. According to the first N-bit address input 01 of the generator, the first group of information inputs 110 (a group of inputs A ₁ -A _N ) of the first selector of the multiplexer 1 is supplied with an N-bit address to which the value of the first element a ₁ should be written. At the control input 07 of the generator, a logic zero value is set, which is fed to input 130 (input SE) of the first selector-multiplexer 1, which leads to switching the address set at input 110 (inputs A ₁ -A _N ) of the first selector-multiplexer 1, to N-bit address input 52 (input for bits A ₁ -A _N ) random access memory 5. M-bit information input 03 of the generator information inputs 53 (inputs D ₁ -D _M ) random access memory 5 is supplied with the value of the first element a ₁ of a given set of data, which is written to the random access memory 5 when it arrives at its inputs 51

and 54

logical zero values for the first inputs "Crystal Select" 04 and "Read/Write" 02, respectively. Then, according to the first N-bit address input 01 of the generator, the first group of information inputs 110 (a group of inputs A ₁ -A _N ) of the first selector of the multiplexer 1 is supplied with an N-bit address, at which the value of the second element a ₂ should be written, and by M - bit information input 03 of the generator to the information inputs 53 (inputs D ₁ -D _M ) of the random access memory 5 is supplied with the value of the element a ₂ and, by setting the values of logical zero at the first inputs "Crystal selection" 04 and "Read / write" 02, write the value of a ₁ in the RAM 5. Similarly, in the RAM 5 are entered all K values of the elements of a given data set. After that, at the first input "Read/Write" 02 of the generator, the value of the logical unit is set.

и 64_k

каждого k-ого блока хранения границ интервалов 6_k. По окончании записи начальных количественных значений верхних границ интервалов в соответствующие блоки хранения границ интервалов, на вторых входах «Выбор кристалла» 08 и «Чтение/запись» 010 устройства устанавливают значение логического нуля.The second step of preparing the generator for operation is as follows. On K P-bit information inputs "Upper limit" 09 ₁ -09 _K generator set the initial quantitative values of the upper boundaries of the intervals. At the same time, the input 09 ₁ is set to the value b ₁ , which is fed to the group of information inputs 63 ₁ (group D ₁ -D _P ) of the block for storing the boundaries of the intervals 6 ₁ , to the input 09 ₂ - the value b ₂ , which is supplied to the group of information inputs 63 ₂ blocks storage boundaries intervals 6 ₂ to the input 09 _K - the value of b _K , which is supplied to the group of information inputs 63 _K block storage boundaries intervals 6 _K . To record the initial quantitative values of the upper limits of the intervals in the storage blocks of the boundaries of the intervals 6 ₁ -6 _K , at the second inputs "Crystal selection" 08 and "Read / write" 010 of the device, the value of the logical unit is set, which is fed to the inputs 61 _k

and _64k

each k-th block of storage of boundaries of intervals 6 _k . At the end of the recording of the initial quantitative values of the upper limits of the intervals in the corresponding storage blocks of the boundaries of the intervals, the second inputs "Crystal selection" 08 and "Read/Write" 010 of the device are set to a logical zero value.

The third step of preparing the generator for operation is as follows. According to the second N-bit address input 05 of the generator, the first N-bit input 21 (inputs of bits A ₁ -A _N ) of the second selector of the multiplexer 2 is supplied with an N-bit address, at which the value of the first element should be written, which determines the new quantitative value of the upper limit given (first) interval. Similarly, in the second selector-multiplexer 2, all K new N-bit values are entered quantitatively (at the first stage of the generator operation) of the specified upper bounds of the intervals. At the control input 07 of the generator, a logical zero value is set, which is fed to input 23 (input SE) of the second selector-multiplexer 2, which leads to switching the address written in binary code and set at input 21 of the second selector-multiplexer 2 to an N-bit information input 31 (input registers D ₁ -D _N ) of the first register 3. In the mode of preparing the generator for operation, the reset inputs 32 and 102 (inputs R) of the first 3 and second 10 registers are respectively fed with the values of a logical unit.

The second N-bit information input 22 (input for bits B ₁ -B _N ) of the second selector-multiplexer 2 from the output 104 of the second register 10 is supplied with the previous (from the last step) value of the first element with ₁ from the set of values of the probabilities of transitions of the random process from the state to a state that defines a new, quantitatively specified value of the upper limit of the given interval. This value of the first element with ₁ is written to the first register 3, and in order to record new for this step quantitative or verified and confirmed, converted (using NNS) values of the upper limits of the intervals to the first register 3, at the first input "Installation" 06 of the device, the value of the logical unit, which enters the initialization input 33 (input C) of the first register 3. Upon completion of writing to the first register 3 new (for this step) quantitative or verified and confirmed (using NNS) values of the upper limits of the intervals in the first register 3 at the first input "Setting" 06 devices set the value of logical zero.

After the above steps, the generator is ready for operation.

In the generation mode, the operation of the device, taking into account not only the quantitatively set values of the boundaries of the state intervals, but also these values, set taking into account the complex uncertainty, which manifests itself simultaneously and collectively in the form of unreliability and fuzziness of information, occurs as follows (see Fig. 1).

The control input 07 of the generator is supplied with a logical unit value, which is fed to the selection input 13 (input SE) of the first selector-multiplexer 1, which ensures switching of the address coming from the output 104 of the second register 10 to the N-bit address input 52 (input for bits A ₁ -A _N ) random access memory 5.

The source of random numbers 4 in the presence of a logical unit value at the control input 07 of the generator generates a P-bit random address value, which is fed simultaneously to the P-bit inputs "Random number" 71 ₁ -71 _K K comparison blocks 7 ₁ -7 _K , where comparison of the random value of the address with the initial quantitatively specified values of the upper limits of the specified intervals B={b ₁ , b ₂ , ..., b _K } (see Fig. 4). If the received address value belongs to the k-th interval, i.e. it is less than or equal to the quantitatively specified value of the upper limit of the k-th interval, then at the output "Inequality" (output A> B) of the comparators 7.1 ₁ -7.1 _k-1, the values of a logical unit are formed, and at the outputs "Inequality" of the remaining comparators, the values of a logical zero .

) шифратора приоритетов 8.The signals from the "Inequality" outputs of each comparator 7.1 ₁ -7.1 _K through the OR elements 7.3 ₁ -7.3 _K and the outputs "Comparison result" 73 ₁ -73 _K of the comparison blocks 7 ₁ -7 _K are fed to the corresponding inverse inputs 81 ₁ -81 _K (register inputs

) priority encoder 8.

) шифратора приоритетов 8 будут установлены значения логической единицы, а на входах 81_k-1-81_K (входах регистров

) - значения логического нуля. В этом случае на N инверсных выходах 82₁-82_N (выходах

) шифратора приоритетов 8 будет сформирован двоичный код в инверсном представлении, соответствующий значению первого номера входа

с установленным значением логического нуля, т.е. код соответствующий числу k - номеру интервала, которому принадлежит значение a _k. Полученный код после инвертирования в элементах 9₁-9_N поступает на N-разрядный информационный вход 101 второго регистра 10, здесь регистрируется как код предыдущего шага, затем поступает на второй N-разрядный информационный вход 120 (вход для разрядов B₁-B_N) первого селектора-мультиплексора 1 и второй N-разрядный информационный вход 22 (вход для разрядов B₁-B_N) второго селектора-мультиплексора 2.Thus (see Fig. 1), under the condition b _k-1 < a _k < b _k , at the inputs 81 ₁ -81 _K (inputs of the registers

) priority encoder 8 will be set to the values of a logical unit, and at the inputs 81 _k-1 -81 _K (inputs of the registers

) - values of logical zero. In this case, on N inverted outputs 82 ₁ -82 _N (outputs

) of the priority encoder 8, a binary code will be generated in the inverse representation corresponding to the value of the first input number

with the set value of logical zero, i.e. code corresponding to the number k - the number of the interval to which the value a _k belongs. The resulting code, after inversion in elements 9 ₁ -9 _N , is fed to the N-bit information input 101 of the second register 10, is registered here as the code of the previous step, then enters the second N-bit information input 120 (input for bits B ₁ -B _N ) the first selector-multiplexer 1 and the second N-bit information input 22 (input for bits B ₁ -B _N ) of the second selector-multiplexer 2.

From the N-bit output 24 of the second selector-multiplexer 2, the previous (from the last step) values of the upper limits of the intervals, identified both quantitatively and at the same time unreliably and fuzzy, are fed to the N-bit direct input 111 of the neuro-fuzzy network block 11 for verification and identifying the values of these upper limits of the intervals, set and identified taking into account the complex uncertainty, as well as for processing and transforming such initial data, given simultaneously in an unreliable and fuzzy form, to a form that allows to reliably and unambiguously identify and interpret the values of these initial data on the upper limits intervals for different states of the process under study.

In addition to data on the initial and current values of the upper limits of the intervals obtained at the N-bit output 24 of the second selector-multiplexer 2 based on the P-bit random address value generated by the random number source 4 (as provided in the prototype device), and which can be identified both quantitatively and sometimes unreliably, as part of the dynamic control of the device operator by the boundaries of the intervals, it is possible to introduce external control actions, i.e., code values characterizing the new values of the upper limits of the intervals required by the operator. This happens by external input of an N-bit code characterizing these values of the upper limits of the intervals, newly introduced in the control dynamics, through the N-bit neuro-fuzzy input 114 of the block of the neuro-fuzzy network 11 from the N-bit neuro-fuzzy input 014 of the device (see Fig. 1).

This implies that the new values of the upper limits of the intervals introduced by the operator from the outside of the control are guaranteed to need verification and confirmation, that these values a priori have a complex uncertainty, that the new threshold values of the states introduced by the operator for a particular step of modeling discrete random processes are set both unreliably and fuzzy .

Implementation of verification and confirmation procedures using procedures for neuro-fuzzy verification of the values of the upper limits of the intervals that are set and identified taking into account complex uncertainty, as well as procedures for mathematically correct processing and transformation of such initial data, given simultaneously in an unreliable and fuzzy form, to a form that allows reliable and to uniquely identify and interpret them, is carried out in the block of the neuro-fuzzy network 11 as follows.

From the N-bit output 24 of the second selector-multiplexer 2, the previous (from the last step) values of the upper limits of the intervals, identified both quantitatively and at the same time unreliably and fuzzy, arrive and are written in binary code through the N-bit direct input 111 to N- bit input 11.1-1 of the counter 11.1 of the neuro-fuzzy network block 11.

The neuro-fuzzy network block 11 can be implemented in accordance with the scheme proposed in FIG. 2. Sequential comparison by the number of digits of the input data in binary code - the values of the upper limits of the intervals, identified both quantitatively and at the same time unreliable and fuzzy, and making a decision about their mathematical nature - the values of the upper limits of the intervals are set parametrically, quantitatively or have a complex uncertainty and are given, in our case, by the weights of connections for neurons of the first layer of the NNS (acting as parameters of the membership functions of fuzzy sets for the implementation of the rules of the NNS), is carried out in the counter 11.1 and the storage register 11.2 of the block of the neuro-fuzzy network 11 as follows.

Initially, information, i.e., quantitative data and data on the initial and current values of the upper limits of the intervals, set qualitatively - both unreliable and fuzzy, coming from the N-bit output 24 of the second selector-multiplexer 2, differs in the number of bits: for recording in In the binary code of quantitative information, 5 (five) digits of the binary code are sufficient, while information with complex uncertainty carries, in addition to the usual number, the characteristic of connection weights for neurons of the first layer of the NNN (parameters of membership functions), which objectively requires the use of at least 10 ( ten) bits of a binary code for recording and storing neuro-fuzzy information - both unreliable and fuzzy given values of the upper boundaries of the intervals. Taking into account this fact, the counter 11.1 and the storage register 11.2 of the block of the neuro-fuzzy network 11 are built.

The counter 11.1 and the storage register 11.2 of the block of the neuro-fuzzy network 11 are designed to store five digits of incoming information, if the number of digits exceeds this figure, then, from the point of view of mathematics, this information is the initial and current step-by-step values of the upper limits of the intervals for the states of the generated random process, arrives in a qualitative form - they have both unreliability and fuzziness. In this case, both the counter 11.1 and the storage register 11.2 of the block of the neuro-fuzzy network 11 perform the functions of a transit node, moreover, from the N-bit information output 11.2-1 of the storage register 11.2, this information in binary code immediately arrives for additional verification and confirmation using procedures neuro-fuzzy verification of the values of the upper limits of the intervals, set and identified taking into account complex uncertainty, to the N-bit input 11.3-1 (I ₁ -I _N ) of the neuro-fuzzy programmable calculator 11.3 (see Fig. 2).

Also, to the N-bit input 11.3-1 (I ₁ -I _N ) of the neuro-fuzzy programmable calculator 11.3 through the N-bit neuro-fuzzy input 114 of the block of the neuro-fuzzy network 11 from the N-bit neuro-fuzzy input 014 of the device within the dynamic control of the device operator by the boundaries of the intervals, external control actions are received in a binary code, i.e., the values of the code characterizing the new values of the upper limits of the intervals required by the operator. These values (the values of the upper bounds of the intervals) are a priori set taking into account the complex uncertainty and need additional verification and confirmation using neuro-fuzzy verification procedures.

If the N-bit direct input 111 and the N-bit input 11.1-1 of the counter 11.1 of the block of the neuro-fuzzy network 11 receive information in binary code (the values of the upper boundaries of the state intervals of the generated random process) in the amount of five bits, then this information enters unambiguous, reliable form, has a quantitative meaning and through the N-bit output 11.1-2 of the counter 11.1 enters the N-bit input 11.2-2 of the storage register 11.2. The storage register 11.2 records this information and from its N-bit direct output 11.2-4 through the N-bit direct output 112 of the block of the neuro-fuzzy network 11 sends this data (given in an unambiguous, reliable form, the values of the elements of the generated random process are quantitatively specified, not requiring additional verification and confirmation using procedures of neuro-fuzzy verification of the value of the upper limits of the intervals for a particular step) to the N-bit information input 31 (input of registers D ₁ -D _N ) of the first register 3 (see Fig. 1).

Data characterizing the values of the upper limits of the intervals (for each subsequent step), defined, recognized in the counter 11.1 and stored in the storage register 11.2 of block 11 as both unreliable and fuzzy, and requiring additional verification and confirmation using neuro-fuzzy verification procedures using NNS come from the N-bit information output 11.2-1 of the storage register 11.2 to the N-bit input 11.3-1 (I ₁ -I _N ) of the neuro-fuzzy programmable calculator 11.3 of the block of the neuro-fuzzy network 11 (see Fig. 2).

The neuro-fuzzy programmable calculator 11.3 of the block of the neuro-fuzzy network 11 records, stores the results of the analysis of the code characterizing the values of the upper boundaries of the intervals, and mathematically correct neuro-fuzzy transformation (verification) of the values of this code. Neuro-fuzzy transformation, i.e. additional verification and confirmation with the help of the NNN of the true values of the initial data - the values of the upper limits of the intervals, set simultaneously in an unreliable and fuzzy form to a form that allows you to reliably (completely, consistently) identify and interpret the values of these upper limits of the intervals for a specific step of modeling discrete random processes, is carried out in the neuro-fuzzy programmable computer 11.3 block neuro-fuzzy network 11 as follows (see Fig. 2).

The neuro-fuzzy programmable calculator 11.3 of the block of the neuro-fuzzy network 11 (see Fig. 2) is technically implemented on the basis of a programmable (in terms of the weight matrix of connections for neurons of the first layer of the NNN, which are parameters of membership functions formulated by experts) microprocessor section (for example, commercially available MPS K1804BC1), which acts as a programmable parallel arithmetic logic unit (ALU) that implements a computational neuro-fuzzy algorithm for the operation of a five-layer NNN of direct signal propagation, such as ANFIS (Adaptive-Network-Based Fuzzy Inference System - adaptive fuzzy inference network) and described in [1, 2].

If the information from the N-bit neuro-fuzzy input 014 of the device, from the N-bit neuro-fuzzy input 114 of the neuro-fuzzy network block 11 and from the N-bit information output 11.2-1 of the storage register 11.2 of the neuro-fuzzy network block 11 to N- the bit input 11.3-1 (I ₁ -I _N ) of the neuro-fuzzy programmable calculator 11.3 of the block of the neuro-fuzzy network 11 is not received, respectively, the command is not received, initiating the start of the procedure of neuro-fuzzy conversion of the values of the upper boundaries of the intervals to the input 11.3-2 of the resolution of the outputs A (OE _I ) neuro-fuzzy programmable computer 11.3.

In this case, the N outputs A (A ₁ -A _N ) of the neuro-fuzzy programmable calculator 11.3, which means the N-bit output 11.4-2 of the storage element 11.4 and the N-bit information output 113 of the block of the neuro-fuzzy network 11 are blocked. Otherwise, there is a signal at the input 11.3-2 to enable outputs A (OE _I ) of the neuro-fuzzy programmable calculator 11.3 and at the same time unreliable and fuzzy values of the upper limits of the intervals (for each subsequent step), or from the N-bit information output 11.2-1 register storage 11.2 block neuro-fuzzy network 11 or N-bit neuro-fuzzy input 014 of the device through the N-bit neuro-fuzzy input 114 block neuro-fuzzy network 11 arrive at the N-bit input 11.3-1 (I ₁ -I _N ) neuro-fuzzy programmable computer 11.3, which performs the functions of a programmable parallel ALU and is capable of implementing a computational neuro-fuzzy algorithm for the operation of a five-layer NNS direct signal propagation.

The neuro-fuzzy programmable calculator 11.3 of the block of the neuro-fuzzy network 11 (see Fig. 2), which implements the functions of a programmable parallel ALU, based on the programmed values of the elements of the matrix of connection weights for neurons of the first layer of the NNS - the analytically described parameters of the membership functions, formulated by experts, performs calculation procedure in accordance with the computational neuro-fuzzy algorithm for the operation of a five-layer NNN of direct signal propagation, described in detail in [1, 2].

In this case, the input bits (cells) I ₁ -In of the N-bit input 11.3-1 (input I) correspond to the bit (1, ..., N) of the serial code arriving at the N-bit input 11.3-1 (input I) of the neuro-fuzzy programmable calculator 11.3 and are equal to N inputs of neurons of the first layer of the NNS, which is fed with the values of N bits of the code, which has a physical meaning at the same time unreliable and fuzzy defined values of the upper boundaries of the intervals. A set of direct and feedback NNNs (see Fig. 3), programmatically implemented within the neuro-fuzzy programmable calculator 11.3, allows you to take into account the weight coefficients for neurons of the first layer of the NNN - analytically described parameters of membership functions formulated by experts, and, based on the implementation of neuro -of a fuzzy computational algorithm described by expressions (3)-(9) of this description, to receive at N outputs A (A ₁ -A _N ) of a neuro-fuzzy programmable calculator 11.3 the values of N bits of a parallel code, having the physical meaning of additionally verified and confirmed, mathematically correctly verified value of the upper limits of the intervals determined on the basis of reliable (complete) initial data.

At the same time, by implementing a neuro-fuzzy computational algorithm for the operation of a typical five-layer NNN of direct signal propagation (see Fig. 3), described in detail in [1, 2] and illustrated by expressions (3)-(9) of this description, from the point of view of the theory NNS neuro-fuzzy programmable calculator 11.3 block of neuro-fuzzy network 11 performs arithmetic-logical functions: fuzzification, i.e., transformation of numerical input values in the degree of correspondence to linguistic variables (functions of the first layer of NNS); recording and storing fuzzy rules, which are a set of fuzzy rules of the "If-Then" type (functions of the second layer of the NNN); recording and storing data that determine the membership functions of fuzzy sets used in fuzzy rules (functions of the third layer of the NNN); decision-making, i.e., performing an inference operation based on the existing rules (functions of the fourth layer of the NNN) and defuzzification functions aimed at converting the results of the inference into numerical values of the initial data on the boundaries of the intervals for various states of the process under study (functions of the fifth, output layer of the NNN ). Moreover, the supply to the n-th, where n=1, 2, ..., N, input (I _n ) of the N-bit input 11.3-1 (input I) of the neuro-fuzzy programmable calculator 11.3 of the code bit value characterizing both unreliably and fuzzy identified the values of the upper limits of the intervals (for each subsequent step), initiates the output from the corresponding n-th output (A _n ) of the neuro-fuzzy programmable calculator 11.3 (from the output of the summing output neuron of the fifth layer of the NNN data of the n-th neuron of the fourth, penultimate layer) programmed, according to the computational neuro-fuzzy algorithm described in [1, 2], the value of the additionally verified and confirmed, mathematically correct verified value of the upper bounds of the intervals (Fig. 2).

As a result, at the N outputs A (A ₁ -A _N ) of the neuro-fuzzy programmable calculator 11.3, at the corresponding bits of the N-bit output 11.4-2 of the storage element 11.4 and at the corresponding bits of the N-bit information output 113 of the block of the neuro-fuzzy network 11 (see Fig. 2), we obtain information characterizing (based on the analysis obtained in the framework of the work of the NNS) additionally verified and confirmed, verified, unambiguous and true values of the elements of the generated random process - the values of the upper limits of the intervals (for each subsequent step) in in the interests of increasing the reliability of the implementation of the generated sequence. The storage element 11.4 of the block of the neuro-fuzzy network 11 (see Fig. 2) records, stores and outputs from the corresponding bits of its N-bit output 11.4-2 through the corresponding bits of the N-bit information output 113 of the block of the neuro-fuzzy network 11 to the corresponding bits N-bit information input 31 (input of registers D ₁ -D _N ) of the first register 3 (see Fig. 1) code containing additionally checked and confirmed, verified results of the analysis of the values of the upper limits of the intervals - a code that uniquely (completely) and numerically determines threshold values of the probability of occurrence of the corresponding elements of a given data set.

Thus, at the N-bit information input 31 (input of the registers D ₁ -D _N ) of the first register 3, which is responsible for registering and storing values as initial and quantified (coming from the N-bit direct output 112 of the neuro-fuzzy network block 11 (see Fig. 1)), and new, additionally checked and confirmed, verified with the help of NNS, values of the upper limits of the intervals (coming from the N-bit information output 113 of the block of the neuro-fuzzy network 11), we have reliably, unambiguously and completely certain upper threshold values of the probability of occurrence of the corresponding elements of a given data set for a specific step of the generated random process.

This ensures the generation of a data set, taking into account both the quantitatively specified values of the upper boundaries of the intervals into which the set of addresses of this data set is divided, and taking into account the complex uncertainty of the values of these interval boundaries for various states of the generated random process. Current (changed compared to the initial ones) and additionally verified and confirmed with the help of NNS, reliably, unambiguously, fully defined values of the upper limits of the intervals for a specific step from the N-bit output 34 of the first register 3 are fed to the N-bit address inputs 62 ₁ -62 _K K blocks of storage of boundaries of intervals 6 ₁ -6 _K and further on P-bit inputs "Upper limit" 72 ₁ -72 _K comparison blocks 7 ₁ -7 _K (see Fig. 4).

The R-bit inputs "Random number" 71 ₁ -71 _K of the comparison blocks 7 ₁ -7 _K still receive the P-bit random address value from the source of random numbers 4, again, but for this particular step, the random address value is compared with current verified and confirmed, reliably, unambiguously and fully defined values of the upper limits of the given intervals, obtained taking into account both quantitative and qualitative - both unreliable and fuzzy measure of the complex uncertainty of the threshold values of the probability of occurrence of the corresponding elements of a given data set for a specific step of the generated random process . The cycle repeats, with the values of the previous address coming from the output 104 of the second register 10 to the second N-bit information input 120 (input for bits B ₁ -B _N ) of the first selector-multiplexer 1, then to the N-bit address input 52 (input for bits A ₁ -A _N ) random access memory device 5, serve as the initial data for obtaining subsequent values of the element from a given set of data.

As a result, in accordance with random addresses generated by the source of random numbers and in accordance with those entered from the outside (through the N-bit neuro-fuzzy input 014 of the device) and corrected at each step, verified and confirmed, reliably, unambiguously and fully defined values of the upper limits of the given intervals obtained taking into account both quantitative and qualitative - both unreliable and fuzzy measures of complex uncertainty, the current probabilistic-temporal values of the elements A={ a ₁ , a ₂ , ..., a _K ] of the given data set are read from random access memory 5 , which are fed through the output 55 to the M-bit output "Result" 013 of the generator.

) оперативного запоминающего устройства 5 подают значение логической единицы.The generator stops working when its control input 07 is supplied with a logical zero value, which corresponds to the termination of the formation of random addresses by the source of random numbers, or when the first input "Crystal Selection" 04 of the generator is sent to input 51 (input

) random access memory 5 serves the value of a logical unit.

Thus, the claimed device provides an increase in the reliability of the generated random sequence by providing the possibility of generating a given data set, taking into account both the quantitatively specified values of the upper boundaries of the intervals into which the set of addresses of this data set is divided, and taking into account the complex uncertainty of the values of these interval boundaries for various states of the generated random process. The implementation of the possibility of generating a random sequence, taking into account the presence of not only quantitatively, but also at the same time unreliably and fuzzy set values of these upper limits of the intervals, occurs due to the additional verification and confirmation implemented in the block of the neuro-fuzzy network 11 using the procedures of neuro-fuzzy verification of the values of the upper limits of the intervals given and identified taking into account the complex uncertainty, as well as mathematically correct processing and transformation of such initial data, given simultaneously in an unreliable and fuzzy form, to a form that allows them to be reliably and unambiguously identified and interpreted in the interests of implementing the procedure of parametric modeling (generation) of random processes.

This result is due, as a result, to the receipt at the M-bit output "Result" 013 of the generator of the current probabilistic-temporal values of the elements of a given data set, taking into account the presence of both quantitatively given values of the upper boundaries of the intervals into which the set of addresses of this data set is divided, and with taking into account the complex uncertainty of the values of these interval boundaries for various states of the generated random process

An analysis of the principle of operation of the claimed random sequence generator shows the obviousness of the fact that, along with the possibilities stored and described in the prototype for generating a sequence of a given data set, taking into account the probabilistic relationships between the states of this process, the random sequence generator is capable of generating with high reliability the values of the elements of a random process characterizing the real behavior of a complex computing system - when the upper boundaries of the intervals for various states of this random process have both quantitatively and qualitatively - both unreliable and indistinctly expressed physical meaning.

This random sequence generator provides an increase in the level of reliability of modeling (generation) of random processes occurring in real systems, where both Markov models and poorly formalized models are widely used, observed under conditions of complex uncertainty, characterized and described at the same time unreliably and fuzzy given parameters, which expands the functionality of universal generating devices in modern computing, where the claimed random sequence generator will be used.

INFORMATION SOURCES

1. R. Fuller. Introduction to Neuro-Fuzzy Systems. Advances in Soft Computing Series. Springer - Verlag, Berlin, 1999. - 289 p.;

2. Khizhnyakov Yu.N. Fuzzy, neural and neuro-fuzzy control algorithms in real-time systems. - Perm: PNIPU, 2013. - 160 p.;

3. Parashchuk I.B., Mikhailichenko N.V. Features of the use of neuro-fuzzy models for decision support systems in the tasks of evaluating the effectiveness of the functioning of specialized data centers // Information and space. No. 1, 2019. S. 84-88;

4. Andrievskaya N.V., Reznikov A.S., Cheranev A.A. Features of the use of neuro-fuzzy models for the problems of synthesis of automatic control systems. // Fundamental research. Technical science. No. 11. - 2014. S. 1445-1449;

5. Rogozin O.V. Neuro-Fuzzy Adaptive System for Supporting Innovative Decision Making. // Educational technologies. No. 3. - 2011. S. 100-110;

6. Nesteruk G.F., Kupriyanov M.S., Elizarov S.I. To the solution of the problem of neuro-fuzzy classification. // Proceedings of the VI Intern. conf. SCM'2003. - St. Petersburg: SPGETU, 2003. Vol. 1. S. 244-246;

7. Ivanenko V.I., Labkovsky V.A. The problem of uncertainty in decision making problems. - Kyiv: Science thought. 1990. - 136 p.;

8. Parashchuk I.B., Bashkirtsev A.S., Mikhailichenko N.V. Analysis of Levels and Types of Uncertainty Affecting Decision-Making in Information Systems Management // Information and space. No. 1, 2017. P. 112-120;

9. Bayuk O.A., Brailov A.V., Denezhkina I.E. et al. Acceptance of financial decisions under comparative uncertainty. - M.: High school textbook. 2014. - 106 p.;

10. Military systems engineering and system analysis. Models and methods of preparing and making decisions in complex organizational and technical complexes under conditions of uncertainty and multi-criteria. / Ed. B.V. Sokolov. - St. Petersburg: VIKU them. A.F. Mozhaisky, 1999. - 496 p.;

11. Modeling of information systems. / Ed. O.I. Shelukhin. Tutorial. - M.: Radio engineering, 2005. - 368 p.;

12. Nikitin B.Ya., Skrebkov V.N. Theoretical foundations of computer technology. Part 2, - L: LVVIUS, 1988, S. 9-11;

13. Tikhonov V.I., Mironov M.A. Markov processes - M.: Soviet radio, 1977. - 488 p.

Claims (2)

1. A random sequence generator containing K blocks of comparison (7 ₁ -7 _K ), where K≥2 is the maximum possible power of a given set of generated values, the first selector-multiplexer (1), the second selector-multiplexer (2), the first register ( 3), random access memory (5), K storage blocks of interval boundaries (6 ₁ -6 _K ), priority encoder (8), N inverters (9 ₁ -9 _N ), where N=[log ₂ K] is the number of binary bits sufficient to address the elements of a given set of generated values, the second register (10), the source of random numbers (4), P-bit, where P>N is the bit width of the information inputs of the generator, the output "Random number" (42) is connected to P - bit inputs "Random number" (71 ₁ -71 _K ) K blocks of comparison (7 ₁ -7 _K ), P-bit output (65 _k ) of the k-th storage block of interval boundaries (6 _k ), where k=1, 2, ..., K, is connected to the P-bit input "Upper limit" (72 _k ) of the k-th comparison block (7 _k ), the output "Comparison result" (73 _k ) of the k-th comparison block (7 _k ) is connected to k-th inverse input (81 _k ) of the priority encoder (8), n-th inverse output (82 _n ) of which, where n=1, 2, ..., N, is connected to the input (91 _n ) of the n-th inverter (9 _n ), the selection input (130) of the first selector-multiplexer (1) is connected to the control input (41) of the random number source (4) with the selection input (23) of the second selector-multiplexer (2) and is the control input (07) of the generator, the first N-bit information input (110) of the first selector-multiplexer (1) is the first N-bit address input (01) of the generator, the N-bit output (140) of the first selector-multiplexer (1) is connected to the N-bit address input ( 52) random access memory (5), the inverse inputs "Read/Write" (54) and "Crystal Select" (51) are respectively the first input "Read/Write" (02) and the first input "Crystal Select" (04) generator, and M-bit, where M≥2 is the number of bits sufficient to represent the maximum value from a given set of generated values, the output (55) of the random access memory (5) is the M-bit output "Result" (013) of the generator, M -bit information input (53) of the random access memory (5) is an M-bit information input (03) of the generator, while the P-bit information input (63 _k ) of the k-th interval boundary storage block (6 _k ) is the k-th P-bit information input (09 _k ) of the generator, the n-th digit of the N-bit information input (101) of the second register (10) is connected to the inverse output (92 _n ) of the n-th inverter (9 _n ), and the N-bit output (104) of the second register (10) is connected to the second N-bit information input (120) of the first selector-multiplexer (1) and to the second N-bit information input (22) of the second selector-multiplexer (2), the first N-bit input (21) of which is the second N-bit address input (05) of the generator, the N-bit output (34) of the first register (3) is connected to the N-bit address inputs (62 ₁ -62 _K ) K interval boundary storage blocks (6 ₁ -6 _K ), the corresponding inverted inputs "Crystal selection" (61 ₁ -61 _K ) and "Read / write" (64 ₁ -64 _K ) which are combined and are respectively the second input "Crystal selection" (08) and the second input "Read/Write" (010) of the generator, initialization inputs (33) and (103) of the first (3) and second (10) registers are respectively the first (06) and second (012) inputs "Installation" of the generator, and the reset inputs (32) and (102) of the first (3) and second (10) registers are combined and are the input "Zeroing" (011) of the generator, characterized in that a block of neuro-fuzzy network (11) is additionally introduced, designed to carry out verification procedures and confirmation using procedures of neuro-fuzzy verification of the values of the upper boundaries of the intervals of the states of the generated random process, given and identified taking into account the complex uncertainty, as well as mathematically correct processing and transformation of such initial data, given simultaneously in an unreliable and fuzzy form, to a form that allows reliable and unambiguous identify and interpret them, moreover, the N-bit output (24) of the second selector-multiplexer (2) is connected to the N-bit direct input (111) of the neuro-fuzzy network block (11), the N-bit direct output (112) of which is connected to N-bit information output (113) of the neuro-fuzzy network block (11) and is connected to the N-bit information input (31) of the first register (3), N-bit neuro-fuzzy input (114) of the neuro-fuzzy network block (11 ) is an N-bit neuro-fuzzy input (014) of the generator. 2. The random sequence generator according to claim 1, characterized in that the block of the neuro-fuzzy network (11) consists of a counter (11.1), a storage register (11.2), a neuro-fuzzy programmable calculator (11.3) and a storage element (11.4), The N-bit output (11.4-2) of which is the N-bit information output (113) of the block of the neuro-fuzzy network (11), the N-bit input (11.3-1) of the neuro-fuzzy programmable calculator (11.3) is connected to the N-bit information output (11.2-1) of the storage register (11.2) and is an N-bit neuro-fuzzy input (114) of the neuro-fuzzy network block (11) and an N-bit neuro-fuzzy input (014) of the generator, input (11.3-2 ) permission outputs A of the neuro-fuzzy programmable computer (11.3) is connected to the enabling output (11.2-3) of the storage register (11.2), N outputs A (A ₁ -A _N ) of the neuro-fuzzy programmable computer (11.3) are connected to the corresponding N inputs (11.4-1 ₁ -11.4-1 _N ) storage element (11.4), N-bit direct output (11.2-4) of storage register (11.2) is N-bit direct output (112) of neuro-fuzzy network block (11), The N-bit input (11.2-2) of the storage register (11.2) is connected to the N-bit output (11.1-2) of the counter (11.1), the N-bit input (11.1-1) of which is the N-bit direct input (111) of the block neuro-fuzzy network (11).

Images