Geoscience Reference
In-Depth Information
the classical procedure which is shown in Fig.
3.1
. According to these
gures, the
algorithmic load of the sequential procedure changes dynamically, while at the same
time the classical procedure realizes the data processing stage only on the
finishing
step of the experiment. Hence, the synthesis of the ef
cient decision making system
(DMS) poses the following problems:
i. selection of the criterion for the parameters estimation;
ii. revealing of the probabilistic characteristics of the process studied;
iii. a priori estimation of the possible losses concerning the precision of the
decisions taken; and
iv. prognosis of the dynamic stability for the experiment results.
The DMS should have a wide spectrum of functions:
i. visualization of the measurement data in the form of direct soft copy, discrete
distribution and statistical parameters;
ii. calculation of the statistical characteristics (mean, central second and third order
moments, asymmetry and excess coef
cients, expression of the entropy, etc.);
iii. synthesis of the empirical and theoretical distribution functions;
iv. valuation of the parameters which are used in the Neyman-Pearson and the
sequential procedures of the hypotheses decision; and
v. realization of the user access to all the functions of the decision making system.
According to the scheme represented in Fig.
3.5
the decision-making system
should have an expert control level. The unit DMSP-I controls the decision making
procedure through its inputs and outputs. According to the functions of the sub-
units described in Table
3.1
the user can promptly interfere in any arbitrary stage of
the computer experiment, correcting the parameters of the decision making pro-
cedure or even ceasing it.
The sub-unit CTT manages the calculation process taking into account the
character of the task. It forms the variants that correspond to the concrete combi-
nation of the errors of the
. Based on this combination,
the sub-unit CTT produces a set of parameters to manage the other sub-units.
Depending on
first and second kind,
a
and
b
ed procedures are possible. For example, two
variants for the asymmetric thresholds A and B are:
a
and
b
, the simpli
(1) B=
ʲ∕
(1
− α
)
0, A =(1
− ʲ
)/
α
→
const; and
→
(2) B =
ʲ∕
(1
− α
) = const, A =(1
− ʲ
)/
α
→
∞
.
In other words, the errors
a
and
b
are unequal in value. Particularly:
(1)
ʲ
→
0,
α
= const; or
(2)
ʲ
= const,
α
→
0.
Search WWH ::
Custom Search