Geology Reference
In-Depth Information
vector x, returns 0, 1, or 2 (here o, *, or [], respectively); i.e., it predicts the class
that x belongs to. Hence, the classi
nes some inter-class boundaries in the
input space: Fig.
8.2
also shows a possible set of such boundaries. A SVM classi
er de
er
uses the optimal separating hyperplane as the decision boundary between classes in
the input space. The SVM hyperplane is positioned to give the largest margin
between itself and the nearest classi
cation points.
8.2 Statistical Blockade Steps
In this section the basic steps of SB are illustrated. Assume that we have 1,000
samples from a given measurement (i.e., the size of training sample is n
0
. The size
of size synthetic data is n (e.g., 10,000), and the percentage P
t
(e.g., 99 %) and P
c
(e.g., 97 %), where P
t
and P
c
are the two thresholds. The Algorithm (Fig.
8.3
) of the
SB is used as below.
Algorithm 1: Statistical Blockade method
1: Assume: training sample size
(e.g.,
); total sample size
(e.g.,
)
2:
3:
4:
5:
6:
7:
8:
9:
