How clustering analysis improve neural networks performance ?

Hello,

I divided my training set using clustering analysis into several subsets in order to improve performance. But, i don't know why clustering improve accuracy ?

Answer 1

[Greg Heath]

Clustering is pivotal in designing more accurate elliptical basis function classifiers.

In particular, this allows centering, shaping and modifying clusters based on irregularly obtained measurements instead of blindly establishing rectangular grids with fixed width spheres.

I used clustering extensively when designing and modifying EBF classifiers for radar and optical target detectors that were used by the military.

Thank you for formally accepting my answer

Greg

[Greg Heath]

WHOOPS!

I accidently erased your comment wich was incorrectly placed in an ANSWER box.

Sorry.

Greg

[Hamza Ali]

No problem. Thank you for your answer.