Computational Intelligence Techniques - Computational Intelligence in Biomedical Engineering - page 128

c. Bridge of trapezoidal-type function .

⎧
⎨ 1 if s i >a

s i − c

m i = if c ≤ s i ≤ a

(3.85)

⎩ a − c

0 if s i ≤ c

d. Probit function .

m i = φ s i −

µ

σ

where φ ( ) is the cumulative normal distribution function with mean

µ and standard deviation σ .

·

Fuzzy SVMs have recently been applied to problems such as content-based
image retrieval (Wu and Yap, 2006), biomedical analysis (Chen et al., 2005),
and finance (Wang et al., 2005; Bao et al., 2005). The results appear promising
but this hybrid system has several drawbacks, which require further research.
The selection of the membership function and the allocation of the member-
ship grades have to be carefully done to obtain good classification results.
This preprocessing step could make this technique less appealing for online
applications since an exhaustive trial of the membership functions has to be
done. In addition, the capability of the fuzzy SVM to outperform a normal
SVM operating with optimal parameters still remains to be seen. The bilat-
eral fuzzy SVM is slow to train due to the size of the data set and giving only
a small improvement to classification accuracy compared to other methods
such as logistic regression and neural networks (Wang et al., 2005).

3.7 Concluding Remarks
In this chapter, we examined CI techniques that have been successfully applied
to solve various problems. We have concentrated on ANN, SVM, HMMs,
and fuzzy logic systems because of their frequent application in the field of
biomedical engineering. We introduced these methods and discussed the key
issues of their formulation and implementation. In the following chapters,
the application of these techniques to the processing of data for diagnosis of
various diseases will be described.

References
Abe, S. (1997). Neural Networks and Fuzzy Systems: Theory and Applications .
Boston, MA: Kluwer Academic.