Train the classificator with one data set.
learn_sampset_box trains the classificator with data for the key SampKey (see read_sampset). The training sequence is terminated at least after NSamples examples. If NSamples is bigger than the number of examples in SampKey, then a cyclic start at the beginning occurs. If the error underpasses the value StopError, then the training sequence is prematurely terminated. StopError is calculated with N / ErrorN. Whereby N significates the number of examples which were wrong classified during the last ErrorN training examples. Typically ErrorN is the number of examples in SampKey and NSamples is a multiple of it. If you want a data set with 100 examples to run 5 times at most and if you want it to terminate with an error lower than 5%, then the corresponding values are NSamples = 500, ErrorN = 100 and StopError = 0.05. A protocol of the training activity is going to be written in file Outfile.
|
ClassifHandle (input_control) |
class_box -> integer |
| Classificator's handle number. | |
|
SampKey (input_control) |
feature_set -> integer |
| Number of the data set to train. | |
|
Outfile (input_control) |
filename -> string |
| Name of the protocol file. | |
| Default value: 'training_prot' | |
|
NSamples (input_control) |
integer -> integer |
| Number of arrays of attributes to learn. | |
| Default value: 500 | |
|
StopError (input_control) |
real -> real |
| Classification error for termination. | |
| Default value: 0.05 | |
|
ErrorN (input_control) |
integer -> integer |
| Error during the assignment. | |
| Default value: 100 | |
learn_sampset_box returns 2 (H_MSG_TRUE). An exception handling is raised if key SampKey does not exist or there are problems while opening the file.
learn_sampset_box is local and processed completely exclusively without parallelization.
test_sampset_box, enquire_class_box, write_class_box, close_class_box, clear_sampset
test_sampset_box, enquire_class_box, learn_class_box, read_sampset
Tools
