Search this site

Research Topics

 

Surprise Measure


A Bayesian surprise measure is proposed to assist a system model incorporate the knowledge acquired by new observations. Surprise helps differentiate good data (beneficial) from outliers (detrimental), and thus helps to selectively adapt the model parameters.

 

Shape Matching


Several algorithms that provide both rigid and non-rigid point-set registration are described. These algorithms are used for feature extraction, object matching, image registration, and content-based image retrieval.

 

Kernel Methods


Methods that improve performance by first mapping the data into a high dimensional feature space, reproducing kernel Hilbert space. The mapping is non-linear even though the manipulations in feature space are linearly performed. This provides the advantage of solving non-linear problems using linear methods.

 

Other projects


  • Quadratic mutual information (QMI)
  • Correntropy minimum average correlation energy (CMACE)