Clustering and Density Estimation
Cluster-Weighted Modeling
This
last assignment before the final class project was to implement a fitting function
as in Chapter 12 but using cluster-weighted modeling. I took the simplest local
model for the clusters, which is a constant. Depending on your input you may need
to use many clusters. A better model would at least use a linear function. Also
this is a one dimension model, i.e. one input and one output. The power of this
algorithm comes with multiple dimensions. Here are some plots for various fonctions.
Here is the Matlab
code :
cwm.m