Robust on-line neural learning classifier system for data streams

The ever-increasing integration of technology in the different areas of science and industry has fostered practitioners the design of applications that generate stupendous amounts of data on-line (e.g., smart sensors or network monitoring just to mention two real cases). Extracting information from these data is key, in order to gain a better understanding of the processes that the data are describing. However, learning from these data poses new challenges to traditional data mining techniques, which are not designed to deal with data streams: data in which concepts and noise vary over time.

In this regard, I am proud to present the supervised neural constructivist system (SNCS), a neural Michigan-style Learning Classifier System that has been designed to provide a fast reaction capacity and adaptability to the distinct possible changes in concept (concept drifts) with varying noisy inputs. 

Inheriting the intrinsically on-line fashion of Michigan-style learning classifier systems, SNCS evolves a population of multilayer perceptrons that allow the proposed algorithm to function in a variety of problem situations producing accurate classification of data, whether the data are static or in dynamic streams.

The paper is the following:

It can be downloaded from Springer.


Popular posts from this blog

High Performance Computing, yet another brief installation tutorial

A conceptual machine for cloning a human driver

Beyond state-of-the-art accuracy by fostering ensemble generalization