Extended maximum a posteriori-based kernel classification trained by linear programming (MAPLP) with adjustment parameter (MAPLP-P) and difference-type objective function (MAPLP-D).
NopriadiYukihiko YamashitaPublished in: IJCNN (2012)
Keyphrases
- maximum a posteriori
- linear programming
- objective function
- maximum likelihood
- map estimation
- markov random field
- optimal solution
- support vector
- energy function
- image reconstruction
- feature space
- training set
- lp relaxation
- generalized gaussian
- bayesian framework
- prior model
- linear program
- computer vision
- image classification
- feasible solution
- posterior distribution
- dynamic programming
- em algorithm
- support vector machine
- hyperparameters
- machine learning
- graph cuts
- semi supervised
- motion estimation
- cost function
- pairwise
- standard svm
- model parameters are estimated