However, they are only effective in a number of additional properties of the functions fh (xk) y which in practice are often not implemented. Finally, sometimes it is important to not only get a solution to a particular problem for certain go and n, but also to investigate the sensitivity of the decision to change this background, that the use of classical methods is difficult.
We will show how to use the methods DP these difficulties are easily overcome.
We now describe the problem as a model of PD. Intrinsic property of the distribution process between n entities can be considered as n-step process. For a number of k-step, take number, which funds Xk. At step 1 select 1st venture funds chi at the 2nd step - 2nd company to allocate funds from the remaining x2, etc. Obviously, the variables xk, (k = 1, ..., n) can be considered as control variables. The initial state of the system is characterized by the funds go for distribution. Once isolated, X \ is gi = go-X \ means, etc. The values ​​go, gi, ..., In, characterizing the distribution of the balance after the preceding steps will be regarded as the state parameters. Equations of state are equal.

As the first step numbering corresponding number on the 4th step, and ends with the distribution of conditional optimization started with this step, finish building the circuit to the right, adding the company V, is impossible: it would change all the calculations, since the last of the 4th step. Therefore finish the construction of the scheme, with its new facility on the left and providing it with the index 0. Now numbering steps begins with k = 0 (zero-step), and the initial amount of money will be denoted by g_i. Total gross revenue for the five-step process is Zmax = Z0 * (li).
Copyright ©