Stream-Based Extreme Learning Machine Approach for Big Data Problems

Abstract EN

Big Data problems demand data models with abilities to handle time-varying, massive, and high dimensionaldata.

In this context, Active Learning emerges as an attractive technique for the development of highperformance models using few data.

The importance of Active Learning for Big Data becomes more evidentwhen labeling cost is high and data is presented to the learner via data streams.

This paper presents a novelActive Learning method based on Extreme Learning Machines (ELMs) and Hebbian Learning.

Linearizationof input data by a large size ELM hidden layer turns our method little sensitive to parameter setting.

Overfitting is inherently controlled via the Hebbian Learning crosstalk term.

We also demonstrate that asimple convergence test can be used as an effective labeling criterion since it points out to the amount oflabels necessary for learning.

The proposed method has inherent properties that make it highly attractiveto handle Big Data: incremental learning via data streams, elimination of redundant patterns, and learningfrom a reduced informative training set.

Experimental results have shown that our method is competitivewith some large-margin Active Learning strategies and also with a linear SVM.