Maximum Conditional Likelihood via Bound Maximization and the CEM Algorithm

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 andmaximizationprocess is given to speci cally 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 e ciency and regression results superior to EM are demonstrated.

• Type: Conference
• Year: 1998
• Where: NIPS
• Authors: Tony Jebara, Alex Pentland
• Keywords: Conditional Density Estimation, Conditional Expectation Maximization, Expectation Maximization, NIPS 1998, NIPS 2007