Databases Reference
In-Depth Information
7.4 Distributed Multi-Feature Recognition Perspective
Given the extensive capabilities of existing data capture technologies to
retrieve and generate complex data, it is important to consider a multi-feature
approach for pattern recognition. Existing scheme are not able to scale up with
the enormous Internet-scale data. A distributed perspective in implementing
pattern recognition is required. In this chapter, we have discussed a number of
benefits in implementing a distributed pattern recognition scheme, including
the following:
1. A distributed approach for pattern recognition, such as the DHGN, al-
lows more features to be used in the recognition process, e.g., by allo-
cating an additional DHGN network for each feature recognition.
2. The single classifier mechanism of the DHGN can be used for any number
of features. In contrast, existing multi-feature schemes merely implement
combined-classifiers for classification.
In this chapter, we have presented a distributed approach for multi-feature
pattern recognition. The implementation of a single-cycle learning DHGN
algorithm for distributed feature analysis on a collaborative computational
network was discussed. The proposed approach implements a single classifier
scheme for different feature sets. This is achieved using a divide-and-distribute
approach on the available features for each data set. The proposed approach
is not affected by the curse of dimensionality, which results from multiple fea-
tures. By allowing features to be added to the analysis using available com-
putational networks, the DHGN approach implements a scalable recognition
scheme.
