SDDECIDE Add decisions to a pipeline PD=SDDECIDE(P) PD=SDDECIDE(P,R) Create a set of operating points manually PD=SDDECIDE('w',WEIGHTS,LIST) PD=SDDECIDE('thr',THR,LIST) INPUT P trained pipeline returning soft outputs R SDROC object WEIGHTS Matrix of class weights (op.points vs classes) THR Vector of thresholds LIST SDLIST object with decision names OUTPUT PD pipeline P with added operating point DESCRIPTION SDDECIDE adds an operating point to a pipeline P. If not specified, default operating point is added. For pipelines with single output, zero threshold is used, for two or more outputs, equal weights for all classes. SDDECIDE may also add operating poins from SDROC object R, if provided. SDDECIDE may be used for creating operating point manually if the first argument is a string ('w' for weighting-based or 'thr' for thresholding-based op.points). EXAMPLES Creating decision pipeline with manual operating point: pd=sddecide('w',[0.8 0.2],sdlist('apple','banana')) pd=sddecide('thr',0,sdlist('target','non-target')) SEE ALSO SDCONVERT, SDOPS
sddecide
is referenced in examples:
- kb27: Upgrading to perClass 4
- kb26: Useful tips for confusion matrices
- kb24: Example on building an image detector
- kb23: Example on image classification
- kb22: Note on decision tree performance and speed
- kb19: PRTools compatibility
- kb18: How to protect a trained discriminant against outliers?
- kb17: How to optimize three-class classifier in imbalanced problems
- kb16: Visualize the effect of a change of parameters in a trained classifier
- kb11: Hierarchical classifier: How to build detector-classifier cascade?
- kb7: How to convert LIBSVM Support Vector machine into a pipeline?
- kb3: Perform leave-one-out evaluation
- kb1: How to make decisions at a default operating point?