Valiev V.M. STOCHASTIC SIMULATION AND OPTIMIZE RISK DECISION MAKING OF RESOURCES OF HYDROCARBON DEVELOPMENT

Yu.A. Gadzjiev BINARY RANK SCHEME FOR DEPENDED SOURCES

Let be a discrete sequence of bits, where . Let also be probability dependents the value of bit from previous one bits of stream in such mean, that conditional probability of some value of - th stream bit much exceed the probability . Marked this describe as condition (1).

Now I suggest the approach to adaptive dynamic encoding bit streams as follow. Take a - dimensional binary array , which size is and every index from indexes is binary variable . Then initialize the array some a determined value - for example that be null. Then on every step of the algorithm, by the next bit of stream getting, compare one with the element of array and by dependent of the result this operation doing either

1. 
   writing in output stream binary null if one is same;
   

2. 
   change the value that element on opposite value and then writing in output stream binary unit.
   

From condition (1) follow, that by described above coding in output stream beginning a certain moment become prefer number null bits in comparison to unit. By this, through all coding sequenced is alternating chain null-unit bits, thus for compressing must be applied coding of lens alternating bits series with arbitrarily chosen binary code, for which length of -th codeword is monotonic increasing function of the natural value of .

In the test version of implementation described above scheme for program of coding the computer files been applied code of C(0) the parametrical codeword class (see Yu.A. Gadzjiev “About redundancy of coding in parametrical adaptive ranking scheme” in Trans. of DGTU, 1998, pp. 23-26). For this implementation was became the level of compression in means 80% by coding compact exe-files MSC7.0 compiler and nearly 40-60% by coding text or dbf-files. Since its own simply and apply oriented on bits streams this scheme foremost applied to digital channel of data transmitted for different hardware of technical control systems.

L. A. Kuznetsov, M. G. Zhuravlyova THE IDENTIFICATION OF THE CONNECTION BETWEEN RANDOM VALUES HAVING ANY DISTRIBUTION OF PROBABILITIES

The important task of the analysis of industrial systems is the estimation of connection between input and output values, allocation of the most significant input values (factors), which influence on output values (the functions of response). In the majority of existing techniques of the analysis it is supposed, that the values have normal distribution. However density function of the empirical data frequently does not correspond to such supposition. In this case coefficient of correlation can give incorrect results, therefore it is expedient to use other characteristics, attempt to develop which is undertaken in work.

In a basis of research the empirical data are used. Their processing is labour-consuming process requiring many of calculations. Therefore there was an idea to research an opportunity of reception the results by comparing the histograms of input and output random values not carrying out the further calculations. In the present work was executed the analysis of an opportunity of use the conditional probabilities between various groups of random values and criterion of quantity of the information for an estimation of narrowness of connection and choice most significant factors.

As a result of researches was received, that really immediately on samples it is possible to determine most closely connected with function of the response the input factors without the supposition about a normality density functions of researched values. For it in work the special coefficients of identification the connection on conditional probabilities and criterion of quantity of the information are entered and calculated. Their comparison with existing characteristics is given. The examples illustrating efficiency of the approach are resulted.