| PRTools contents |
MDSC
W = MDSC(A,R,CLASSF)
W = A*FDSC([],R,CLASSF)
D = X*W
| Input | |
| A | Dateset used for training |
| R | Dataset used for representation or a fraction of A to be used for this. Default: R = A. |
| CLASSF | Classifier used in dissimilarity space Default LIBSVC([],[],100) |
| X | Test set. |
| Output | |
| W | Resulting, trained feature space classifier |
| D | Classification matrix |
This is a dissimilarity based classifier intended for a feature respresentation. The training set A is used to compute for every class its own eigenspace. All eigenvectors are used. A dissimilarity space is built by the Manhatten (L1, or Minkowsky-1 or city block) distances between training objects A or test objects X and the representation objects R after transformation (i.e. rotation) them to the eigenspace of the class of the particular represention object.
Note that Euclidean distances are not affected by rotation, but Manhatten distances are.
New objects in feature space can be classified by D = X*W or by D = MAP(X,W). Labels can be found by LAB = D*LABELD or LAB = LABELD(D).
| PRTools contents |