Broad Learning System and its Structural Variations

Abstract

After a very fast and efficient discriminative Broad Learning System (BLS) that takes advantage of flatted structure and incremental learning has been developed, this talk will discuss mathematical proof of the universal approximation property of BLS. In addition, the framework of several BLS variants with their mathematical modellings are given. The variations include cascade, recurrent, and broad-deep combination that cover existing deep-wide/broad-wide structures. From the experimental results, the BLS and its variations outperforms several exist learning algorithms on regression performance over function approximation, time series prediction, and face recognition databases.