Using Self Organizing Maps to accelerate Optimization

What is it about?

Self Organizing Map (a popular, unsupervised learning algorithm) can run in the background of any (evolutionary) optimization algorithm to accelerate convergence of the algorithm. Depending on the tuning parameters, the same global optimal solution may be obtained in as less as 50% of the function evaluations.

Featured Image

Why is it important?

In a scenario where function evaluations are costly (in terms of time taken to return a function value at a point), saving function evaluations is very valuable and can save hours of simulation time.