|
|
|
|
| Programming mappings |
There are many pre-programmed mappings available in PRTools. They are discussed separately. So for the starting user there is no need to program his own mappings. More advanced users and especially those who develop new algorithms for pattern recognition or machine learning may want to have their own versions of existing routines or to define entirely new ones. They should obey the PRTools rules if it is desirable that they can be handled by the PRTools system of storing mappings in a single variable of the class mapping. This is needed, for instance, when they have to be called by routines like testc, cleval or featself . In some subsections it will be shown how this can be done.
First it should be determined what mapping type is needed:
data-in -> data-out. They depend on user defined parameters.
data-in -> mapping-out in case of training, or data-in -> data-out when the trained mapping is applied.
mappings-in -> mapping out.
The below two tables show routines specifically designed for programming mappings and some higher level routines for handling mappings.
| > Routines facilitating the programming of mappings | |
| setdefaults | substitute default parameter values |
| mapping | low level mapping definition |
| define_mapping | defines a mapping of type fixed, untrained or combiner |
| trained_mapping | definition of a trained mapping |
| mapping_task | retrieves the task of the mapping (definition, training, execution) from input parameters |
| isdataset | tests input parameter on dataset type. |
| isdatafile | tests input parameter on datafile type. |
| islabtype | tests dataset or datafile on type of labeling |
| isvaldfile | tests on dataset validity for some problem: number of classes, numbers of objects. |
| cdats | support routine for checking datasets |
| > Handling mappings | |
| writes in the command window the name, the dimensionalities of the input space and the output space (i.e. numbers of features), the mapping type and the command that executes the mapping. | |
| disp | writes in the command window the results of display and getdata. |
| +W | returns the contents of the data field of the mapping W. This is identical to getdata(W). |
| setbatch | sets a flag for controlling the execution of mappings in batch mode. |
| getbatch | retrieval of the mapping batch mode flag |
| dataset | conversion the axes of an affine mapping into a dataset. |
| isaffine | test on affine mapping. |
| isclassifier | test on classifier mapping. |
| iscombiner | test on combining mapping. |
| isfixed | test on fixed mapping. |
| istrained | test on trained mapping. |
| isuntrained | test on untrained mapping. |
| show | shows the axes of an affine mapping as images in case of a feature space of a dataset with object images. |
For further discussions and examples see to the following subsections.
R.P.W. Duin
, January 28, 2013|
|
|
|
| Programming mappings |