Maximum Conditional Likelihood via Bound Maximization and the CEM Algorithm

Tony Jebara and Alex Pentland

Neural Information Processing Systems 11

We present the CEM (Conditional Expectation Maximization) algorithm as an extension of the EM (Expectation Maximization) algorithm to conditional density estimation under missing data. A bounding and maximization process is given to specifically optimize conditional likelihood instead of the usual joint likelihood. We apply the method to conditioned mixture models and use bounding techniques to derive the model's update rules. Monotonic convergence, computational efficiency and regression results superior to EM are demonstrated.

Keywords: CEM, bound maximization, conditional likelihood, latent variables, mixture models
  Compressed Postscript . PDF . Full list of tech reports