Other Applications: Sequential Dependence Modelling and Data Mining - On Statistical Pattern Recognition in Independent Component Analysis Mixture Modelling - page 167

INTERACTIVITY

Email Access

1

4

5

Chat

Agenda

2

Wo rk g ro u p
documents

For um

Contents

Exer cis e s

News

Achievement

PERSONAL ACTIVITY

3

Fig. 7.11 Outline of learning style dimensions associated with e-learning web activities obtained
by non-parametric ICA

Once the true classes (known or estimated) are available, the class-transition
probabilities can be easily estimated.
Some simulations and a real data analysis case have verified the potential
improvements derived from including sequential information in the ICAMM
algorithms that are used for classification purposes. Further work is required to
develop algorithms which can simultaneously estimate all the model parameters in
an unsupervised framework.
The second application, which consisted of applying the proposed non-
parametric ICA to detect the patterns of learning styles in educational web
activities, produced promising results for a real case with huge historical data.
Modelling learning dimensions as a combination of web event activities enhanced
the detection of the student learning styles. Those results could be used to adapt
teaching methodologies or, in general, to improve the learning system, balancing
distributed passive learning (DPL) and distributed interactive learning (DIL).
Thus, the versatility of these methods has been demonstrated into different
kinds of problems; the first problem where there are hidden variables that model
dynamic dependence among class transitions; and the second problem, where the
proposed non-parametric approach is able to estimate suitable sources for the
analysis of huge data (from which it is normally difficult to derive a parametric
model for the source distributions).

References

1. T.W. Lee, M.S. Lewicki, T.J. Sejnowski, ICA mixture models for unsupervised classification
of non-gaussian classes and automatic context switching in blind signal separation. IEEE
Trans. Pattern Anal. Mach. Intell. 22(10), 1078-1089 (2000)
2. T.W. Lee, M.S. Lewicki, Unsupervised image classification, segmentation, and enhancement
using ICA mixture models. IEEE Trans. Image Process. 11(3), 270-279 (2002)
3. R. Choudrey, S. Roberts, Variational mixture of bayesian independent component analysers.
Neural Comput. 15(1), 213-252 (2002)