cluster-weighted modeling (CWM)

• combination of both model types.

• a tool for nonlinear function approximation (supervised learning) that uses soft weighting of locally generalized linear models.

• a nice level of description, derived from
  • Gaussian mixture models (Jordan ’94, Neal ‘93),
  • graphical models (Buntine ‘96, Jordan ‘99),
  • kernel based linear regression and radial basis function type architectures (Girosi et al. ’95)

Previous slide

Next slide

Back to first slide

View graphic version