Learning ELM network weights using linear discriminant analysis

This work addresses a specific computational bottleneck in ELM training for classification tasks, but it is incremental as it builds on existing methods without introducing a new paradigm.

The authors tackled the problem of determining hidden-to-output weights in Extreme Learning Machines for classification by proposing a linear discriminant analysis-based method as an alternative to the pseudo-inverse approach, resulting in Bayes optimal single point estimates for the weights.