set_regularization_params_ocr_class_mlpT_set_regularization_params_ocr_class_mlpSetRegularizationParamsOcrClassMlpSetRegularizationParamsOcrClassMlp (Operator)

Name

set_regularization_params_ocr_class_mlpT_set_regularization_params_ocr_class_mlpSetRegularizationParamsOcrClassMlpSetRegularizationParamsOcrClassMlp — Set the regularization parameters of an OCR classifier.

Signature

Description

set_regularization_params_ocr_class_mlpset_regularization_params_ocr_class_mlpSetRegularizationParamsOcrClassMlpSetRegularizationParamsOcrClassMlpSetRegularizationParamsOcrClassMlp sets the regularization parameters of the OCR classifier passed in OCRHandleOCRHandleOCRHandleOCRHandleOCRHandle. The regularization parameter to be set is specified with GenParamNameGenParamNameGenParamNameGenParamNamegenParamName. Its value is specified with GenParamValueGenParamValueGenParamValueGenParamValuegenParamValue.

As described at create_class_mlpcreate_class_mlpCreateClassMlpCreateClassMlpCreateClassMlp, it may be desirable to regularize the OCR classifier (i.e., the MLP of the OCR classifier) to enforce a smoother transition of the confidences between the different classes and to prevent overfitting of the OCR classifier to the training data. To achieve this, a penalty for large MLP weights (which are the main reason for very sharp transitions between classes) can be added to the training of the OCR classifier in trainf_ocr_class_mlptrainf_ocr_class_mlpTrainfOcrClassMlpTrainfOcrClassMlpTrainfOcrClassMlp by setting GenParamNameGenParamNameGenParamNameGenParamNamegenParamName to 'weight_prior'"weight_prior""weight_prior""weight_prior""weight_prior" and setting GenParamValueGenParamValueGenParamValueGenParamValuegenParamValue to a value > 0. Furthermore, the regularization parameters can be determined automatically. For details, see set_regularization_params_class_mlpset_regularization_params_class_mlpSetRegularizationParamsClassMlpSetRegularizationParamsClassMlpSetRegularizationParamsClassMlp. If the regularization parameters should be determined automatically, please carefully note the advice in set_regularization_params_class_mlpset_regularization_params_class_mlpSetRegularizationParamsClassMlpSetRegularizationParamsClassMlpSetRegularizationParamsClassMlp on how to select the parameters NumHiddenNumHiddenNumHiddenNumHiddennumHidden of the MLP and 'num_outer_iterations'"num_outer_iterations""num_outer_iterations""num_outer_iterations""num_outer_iterations" and on the memory and runtime implications of automatically determining the regularization parameters on the runtime of the training of the OCR classifier.

Execution Information

• Multithreading type: reentrant (runs in parallel with non-exclusive operators).
• Multithreading scope: global (may be called from any thread).
• Processed without parallelization.

This operator modifies the state of the following input parameter:

• OCRHandleOCRHandleOCRHandleOCRHandleOCRHandle

During execution of this operator, access to the value of this parameter must be synchronized if it is used across multiple threads.

Parameters

Example (HDevelop)

* This example shows how to determine the regularization parameters
* automatically without examining the convergence of the
* regularization parameters.
* Create the OCR classifier.
read_ocr_trainf_names ('ocr.trf', CharacterNames, CharacterCount)
create_ocr_class_mlp (8, 10, 'constant', 'default', CharacterNames, \
                      40, 'none', |CharacterNames|, 42, OCRHandle)
* Set up the automatic determination of the regularization
* parameters.
set_regularization_params_ocr_class_mlp (OCRHandle, 'weight_prior', \
                                         [0.01,0.01,0.01,0.01])
set_regularization_params_ocr_class_mlp (OCRHandle, \
                                         'num_outer_iterations', 10)
* Train the classifier.
trainf_ocr_class_mlp (OCRHandle, 'ocr.trf', 100, 1, 0.01, Error, \
                      ErrorLog)
* Read out the estimate of the number of well-determined
* parameters.
get_regularization_params_ocr_class_mlp (OCRHandle, \
                                         'fraction_well_determined_params', \
                                          FractionParams)
* If FractionParams differs substantially from 1, consider reducing
* NumHidden appropriately and consider performing a preprocessing that
* reduces the number of input variables to the net, i.e., canonical
* variates or principal components.
write_ocr_class_mlp (OCRHandle, 'ocr.omc')



* This example shows how to determine the regularization parameters
* automatically while examining the convergence of the
* regularization parameters.
* Create the OCR classifier.
read_ocr_trainf_names ('ocr.trf', CharacterNames, CharacterCount)
create_ocr_class_mlp (8, 10, 'constant', 'default', CharacterNames, \
                      40, 'none', |CharacterNames|, 42, OCRHandle)
* Set up the automatic determination of the regularization
* parameters.
set_regularization_params_ocr_class_mlp (OCRHandle, 'weight_prior', \
                                         [0.01,0.01,0.01,0.01])
set_regularization_params_ocr_class_mlp (OCRHandle, \
                                         'num_outer_iterations', 1)
for OuterIt := 1 to 10 by 1
    * Train the classifier.
    trainf_ocr_class_mlp (OCRHandle, 'ocr.trf', 100, 1, 0.01, Error, \
                          ErrorLog)
    * Read out the regularization parameters
    get_regularization_params_ocr_class_mlp (OCRHandle, \
                                             'weight_prior', \
                                             WeightPrior)
    * Inspect the regularization parameters manually for
    * convergence and exit the loop manually if they have
    * converged.
    * [...]
endfor
* Read out the estimate of the number of well-determined
* parameters.
get_regularization_params_ocr_class_mlp (OCRHandle, \
                                         'fraction_well_determined_params', \
                                          FractionParams)
* If FractionParams differs substantially from 1, consider reducing
* NumHidden appropriately and consider performing a preprocessing that
* reduces the number of input variables to the net, i.e., canonical
* variates or principal components.
write_ocr_class_mlp (OCRHandle, 'ocr.omc')

Result

If the parameters are valid, the operator set_regularization_params_ocr_class_mlpset_regularization_params_ocr_class_mlpSetRegularizationParamsOcrClassMlpSetRegularizationParamsOcrClassMlpSetRegularizationParamsOcrClassMlp returns the value 2 (H_MSG_TRUE). If necessary, an exception is raised.

Possible Predecessors

create_ocr_class_mlpcreate_ocr_class_mlpCreateOcrClassMlpCreateOcrClassMlpCreateOcrClassMlp

Possible Successors

get_regularization_params_ocr_class_mlpget_regularization_params_ocr_class_mlpGetRegularizationParamsOcrClassMlpGetRegularizationParamsOcrClassMlpGetRegularizationParamsOcrClassMlp, trainf_ocr_class_mlptrainf_ocr_class_mlpTrainfOcrClassMlpTrainfOcrClassMlpTrainfOcrClassMlp

Module

OCR/OCV