classify_class_mlpT_classify_class_mlpClassifyClassMlpclassify_class_mlpClassifyClassMlpClassifyClassMlp (Operator)

Name

classify_class_mlpT_classify_class_mlpClassifyClassMlpclassify_class_mlpClassifyClassMlpClassifyClassMlp — Calculate the class of a feature vector by a multilayer perceptron.

Signature

Description

classify_class_mlpclassify_class_mlpClassifyClassMlpclassify_class_mlpClassifyClassMlpClassifyClassMlp computes the best NumNumNumNumNumnum classes of the feature vector FeaturesFeaturesFeaturesFeaturesFeaturesfeatures with the multilayer perceptron (MLP) MLPHandleMLPHandleMLPHandleMLPHandleMLPHandleMLPHandle and returns the classes in ClassClassClassClassClassclassVal and the corresponding confidences (probabilities) of the classes in ConfidenceConfidenceConfidenceConfidenceConfidenceconfidence. Before calling classify_class_mlpclassify_class_mlpClassifyClassMlpclassify_class_mlpClassifyClassMlpClassifyClassMlp, the MLP must be trained with train_class_mlptrain_class_mlpTrainClassMlptrain_class_mlpTrainClassMlpTrainClassMlp.

classify_class_mlpclassify_class_mlpClassifyClassMlpclassify_class_mlpClassifyClassMlpClassifyClassMlp can only be called if the MLP is used as a classifier with OutputFunction = 'softmax'"softmax""softmax""softmax""softmax""softmax" (see create_class_mlpcreate_class_mlpCreateClassMlpcreate_class_mlpCreateClassMlpCreateClassMlp). Otherwise, an error message is returned. classify_class_mlpclassify_class_mlpClassifyClassMlpclassify_class_mlpClassifyClassMlpClassifyClassMlp corresponds to a call to evaluate_class_mlpevaluate_class_mlpEvaluateClassMlpevaluate_class_mlpEvaluateClassMlpEvaluateClassMlp and an additional step that extracts the best NumNumNumNumNumnum classes. As described with evaluate_class_mlpevaluate_class_mlpEvaluateClassMlpevaluate_class_mlpEvaluateClassMlpEvaluateClassMlp, the output values of the MLP can be interpreted as probabilities of the occurrence of the respective classes. However, here the posterior probability ClassProbClassProbClassProbClassProbClassProbclassProb is further normalized as ClassProbClassProbClassProbClassProbClassProbclassProb = p(i|x)/p(x), where p(i|x) and p(x) are defined as in evaluate_class_gmmevaluate_class_gmmEvaluateClassGmmevaluate_class_gmmEvaluateClassGmmEvaluateClassGmm. In most cases it should be sufficient to use NumNumNumNumNumnum = 1 in order to decide whether the probability of the best class is high enough. In some applications it may be interesting to also take the second best class into account (NumNumNumNumNumnum = 2), particularly if it can be expected that the classes show a significant degree of overlap.

Parallelization

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

Parameters

Result

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

Possible Predecessors

train_class_mlptrain_class_mlpTrainClassMlptrain_class_mlpTrainClassMlpTrainClassMlp, read_class_mlpread_class_mlpReadClassMlpread_class_mlpReadClassMlpReadClassMlp

Alternatives

evaluate_class_mlpevaluate_class_mlpEvaluateClassMlpevaluate_class_mlpEvaluateClassMlpEvaluateClassMlp

See also

create_class_mlpcreate_class_mlpCreateClassMlpcreate_class_mlpCreateClassMlpCreateClassMlp

References

Christopher M. Bishop: “Neural Networks for Pattern Recognition”; Oxford University Press, Oxford; 1995.
Andrew Webb: “Statistical Pattern Recognition”; Arnold, London; 1999.

Module

Foundation