Information Technology Reference
In-Depth Information
the GA execution. In addition, an increase of the number of participants allows us to
provide an external validation set for the CCR at the end of the GA execution.
Moreover, it is possible to improve the GA proposed in this chapter. In fact,
improving genetic operators and testing other evaluation criteria are all paths that
remain to be explored. A
final interesting point concerns the transformation of the
prediction obtained (
) to a probability using
linear discriminant analysis or logistic regression as evaluation functions.
After re
“
normal
”
state of alertness or
“
relaxed
”
ning the proposed method, future work will consider integrating this
approach into a full user-friendly experience, where the mental state of the user
directly in
uences the behavior of the system. One example application, that is
relevant
es the
multimedia content that is presented to the user based on his/hers mental state, to
encourage a more pleasant or useful experience.
to the present collection,
is a system that automatically modi
9.8
Questions
1. What is the shape of the raw data?
2. Can we use directly the raw data to classify them?
3. In general, what are the bene
ts to use a wavelet transform instead of a Fourier
transform?
4. In the work presented in this chapter, could we use a Fourier Transform?
5. How can we be sure that the data we use for learning is relevant?
6. Could you summarize in a few lines the behavior of a GA?
7. Why do we use a GA for the optimization of the frequencies and to select the
best electrode?
8. Why do we use the slope criterion as a feature?
9. Could this method be used to classify other types of mental states?
10. Suppose that you work on data from individuals in two modalities described by
a single variable. Draw a scheme explaining the behavior of the Single Value
Classi
er (SVC) algorithm and implement this algorithm.
Acknowledgments
The authors wish to thank V
é
rane Faure, Julien Clauzel, and Mathieu Car-
pentier, who collaborated as interns in the research team during the development of parts of this
work.
References
Anderson C, Sijercic Z (1996) Classification of EEG signals from four subjects during five mental
tasks. In: Proceedings of the conference on engineering applications in neural networks,
London, United Kingdom, pp 407
414
-
Ben Khalifa K, B
doui M, Dogui M, Alexandre F (2005) Alertness states classification by SOM
and LVQ neural networks. Int J Inf Technol 1:131
é
134
-
Breiman L (2001) Random forests. Mach Learn 45:5
32
-
