MULTI_LABELING

Info on the PRTools multiple labeling system.

This is not a command, just an information file.

In the PRTools datasets the data is stored together with the class labels  for the objects. For the crisp and soft labels, as well as for targets a  system of multiple labels has been constructed by which several label sets,  the corresponding class names, prior probabilities and classification costs  can be stored. There is always one of them active.

This system, available from PRTools4.1 is build on top of the old, single  labling system used until PRTools4.0. It is implemented in such a way  that the existing software is not be disturbed by it. The following new  commands are available for using multiple labeling. Recall that PRTools  datasets do not store the object labels as such, but as a label list  (list of class names), stored in the dataset LABLIST field and by a vector  of LABLIST indices (one index per object) stored in the NLAB field.

• [B,N] = ADDLABELS(A,LABELS,LABLISTNAME)
  This adds an additional set of labels to the dataset A, stores it as a  label list using the given label list name (LABLISTNAME). This name, or the  returned number N may be used to activate the labels in future. It is  also activated now.

• B = CHANGELABLIST(A,LABLISTNAME)
  B = CHANGELABLIST(A,N) Activates a previously stored label list.

• B = DELLABLIST(A,LABLISTNAME)
  B = DELLABLIST(A,N)
  Deletes an existing label list.

• GETLABLISTNAMES(A)
  Returns the names of the stored label lists.

• SETLABLISTNAMES(A,NEW_NAMES,OLD_NAMES)
  Change the names of existing label lists.

• [N,LABLISTNAME] = CURLABLIST(A)
  Returns the number and names of the current label list.

• [B,N] = ADDLABLIST(A,LABLIST,LABLISTNAME)
  This is a low level routine, of interest of PRTools maintenance  programmers only. It organises the storage of information. This routine  just stores a new list of label names and makes space for the NLAB indices. It should thereby by followed by a SETNLAB call. Read the  ADDLABLIST help file for more information on the implementation.

The NLAB dataset field is extended to a MxN array, in which M is the  number of objects and N is the number of label lists. The existing  routines like GETLABELS, SETLABELS, GETNLAB, SETNLAB, GETLABLIST,  SETLABLIST, GETPRIOR, SETPRIOR, GETCOST and SETCOST refer to the current  label list only. Interaction with the multiple label system is done  automatically.