Environmental Decision-Making - New Ecoinformatics Tools in Environmental Science: Applications and Decision-making - page 239

Fig. 3.9 Scheme for realization of parallel-sequential procedure to make a statistical decision

possible to reduce the time of data processing, but in this case the problem of the
data

flow agreement exists.
Two situations are possible:

fl

(1) the system has time to analyze the whole chain of vectors without delay in
individual channels, and
(2) at some moments, a delay occurs in some channels. In a general case, two
types of delay are possible: in time and in volume.

Assume constant delays in time ˄ 1 , … , ˄ k , and delays in volume m 1 , m 2 , … , m k .

In the first case, the process of choosing the vector T b is realized by a channel-to-

channel procedure when the vector components T bj are accumulated with the time
delay intervals

˄ i and then they are transmitted to the chain {T bi }. The number of
delayed estimates of T bi in each channel is random and therefore the buffer memory
of each channel should be reserved. In the second case, a random value is the time
of delay fixed. In both cases, it

is necessary to assess the probability not to miss a real value of the vector T b and to

˄ i (i=1, … , k), and the volumes of the buffers mi i are

find optimal values of ˄ i and m i corresponding to the maximum level of this

probability.
Let the on-board computer memory has M cells. The statistical analyzer realizes
with time the simultaneous parallel selection of components of the ”

variant of the vector parameter b=b * . This is achieved by work of the K consec-
utive devices or by algorithms of selection of component values bi i (i =1,

“ suspected

… ,

“ ”

K) corresponding to the space of anomalous features. To the process of

selecting the anomaly variant, it is possible to create two kinds of constant delays
between the selection devices of the anomalous components: constant delay with
number of

stabilize

and constant delay with time.
The two delay modes allow one to have a more

“ candidates ”

flexible organization of the dis-
tribution of Mmemory elements for storing the intermediate values of the

fl

”
components. Because of the random character of the whole process, the statistical
analyzer is apt to make transient errors due to the over

“ suspected

fl ow of the computer memory