Cheat-sheetSDGAUSS Gaussian model P=SDGAUSS(DATA,options) % train from DATA P=SDGAUSS(MEAN,COV,PRIORS) % create from component parameters P=SDGAUSS(MEAN,[],PRIORS) % create from component parameters INPUT DATA Labeled dataset MEAN SDDATA with component means COV cell array with covariance matrices PRIORS vector with prior per component OUTPUT P Gaussian model per class OPTIONS 'prior' Class priors (default: use priors from the training set) 'no display' Do not show progress of regularization optimization Regularization: 'reg' Automatic regularization 'reg',R Regularization constant added to diagonal 'test',TS Use a test/validation set TS to evaluate regularization Do not split DATA internally. 'tsfrac',F Fraction of data used to validating error (default: 0.2) DESCRIPTION SDGAUSS trains a Gaussian model with full covariance matrix per class. The model may be regularized using 'reg' option by adding a constant to covariances' diagonals. SDGAUSS may be trained on one class data set and used for detection. Alternatively, SDGAUSS can create Gaussian model directly from parameters. If COV is empty, it is intialized to unit covariance matrices. EXAMPLES p=sdgauss(data) % Train gaussian model, no regularization p=sdgauss(data,'reg') % run automatic regularization p=sdgauss(data,'reg',0.01) % regularize by adding 0.01 on cov.diagonal pinit=sdgauss(sddata([0 0; 1 1; 2 2]), [],[0.3 0.3 0.3]) READ MORE http://perclass.com/doc/guide/classifiers.html#sdgauss SEE ALSO SDQUADRATIC, SDLINEAR, SDMIXTURE, SDNMEAN
sdgauss is referenced in examples:
- kb24: Example on building an image detector
- kb18: How to protect a trained discriminant against outliers?
- kb17: How to optimize three-class classifier in imbalanced problems
- kb13: How to find samples with a specific type of error in a confusion matrix?
- kb12: Detector classifier cascade with ROC analysis
- kb11: Hierarchical classifier: How to build detector-classifier cascade?
- kb8: How to build a detector in a single line of code?