addlablist
ADDLABLIST
Add another label list to dataset definition
[B,N] = ADDLABLIST(A,LABLIST,LABLISTNAME)
Input | A | Dataset |
LABLIST | Label list to be added |
LABLISTNAME | Optional name for this label list |
Output | B | Dataset |
N | Number for the new label list |
Description
This command adds an additional label list to a dataset. It also activates this label list as the current one. LABLISTNAME may be given to identify this label list in the CHANGELABLIST command. Alternatively, the label list number may be used there.
Multiple label lists are only supported for crisp label types and not for the soft and target label types. See SETLABELS.
In addition to storing the new label list the following changes in the dataset are made
- If the dataset did not contain a mulitple label setup yet, it is implemented. The following definition is used for L label lists
- B.LABLIST becomes a cell array of size (L+1,2)
- B.LABLIST(I,1) contains label list I
- B.LABLIST(I,2) contains the corresponding prior probabilities (to be substituted in B.PRIOR by CHANGELABLIST)
- B.LABLIST(I,3) contains the corresponding error costs (to be substituted in B.COST by CHANGELABLIST)
- B.LABLIST(I,4) contains the corresponding label type (crisp, soft or targets).
- B.LABLIST(L+1,1) contains a char array with the label list names These names can be retrieved by GETLABLISTNAMES
- B.LABLIST(L+1,2) contains the index of the current label list.
- B.LABLIST(L+1,3) contains a vector with target sizes (number of columns) for each label list. Used by SETTARGETS and GETTARGETS.
- B.NLAB is an array of size (M,L), in which M is the number of objects in B
- The number of columns in NLAB is extended by one. This columns is filled by zeros (no labels defined yet).
- B.PRIOR is made empty
- B.COST is made empty
- An new cell pair is added in B.LABLIST just before the last pair. it is filled by the new label list and an empty prior definition
- The set of label lists in B.LABLIST{L+1,1) is updated with the new label list name.
- The current label list in B.LABLIST{L+1,2) is set to L.
The ADDLABLIST command should be followed by a SETNLAB to store the proper new labels for the objects and by a SETPRIOR to update the prior probability setting. These actions may be combined by the ADDLABELS command. An existing label list may be removed by the DELLABLIST command.
[A,N,T0,T1] = ADDLABLIST(A)
In this case just the current label list is returned in N and the start and endpoint of the corresponding columns in A.TARGET. If T0no targets are set.
See also
datasets, multi_labeling, addlabels, changelablist, curlablist, dellablist,