Objective Function

The objective function of BLS can be flexibly changed rather than fixed. For different application backgrounds or goals, we can make the BLS perform better by modifying the objective function.
At present, [1, 2] has improved the BLS performance by modifying the regular terms in the BLS objective function. Experiments showed that the classification accuracy rate of the regularized BLS are significantly improved and the new developed models are more robust.
In addition, [3] modified the objective function to apply BLS on large-scale chaotic time series prediction problem.

Related articles are listed below.

[1] Discriminative Graph Regularized Broad Learning System for Image Recognition

[2] Regularized Robust Broad Learning System for Uncertain Data Modeling

[3] Structured Manifold Broad Learning System: A Manifold Perspective for Large-Scale Chaotic Time Series Analysis and Prediction