Return a training sample from the training data of a multilayer perceptron.

get_sample_class_mlp reads out a training sample from the multilayer perceptron (MLP) given by MLPHandle that was stored with add_sample_class_mlp. The index of the sample is specified with NumSample. The index is counted from 0, i.e., NumSample must be a number between 0 and NumSamples - 1, where NumSamples can be determined with get_sample_num_class_mlp. The training sample is returned in Features and Target. Features is a feature vector of length NumInput, while Target is a target vector of length NumOutput (see add_sample_class_mlp and create_class_mlp).

get_sample_class_mlp can, for example, be used to reclassify the training data with classify_class_mlp in order to determine which training samples, if any, are classified incorrectly.

* Train an MLP
create_class_mlp (NIn, NHidden, NOut, 'softmax', 'canonical_variates',
                  NComp, 42, MLPHandle)
read_samples_class_mlp (MLPHandle, 'samples.mtf')
train_class_mlp (MLPHandle, 100, 1, 0.01, Error, ErrorLog)
* Reclassify the training samples
get_sample_num_class_mlp (MLPHandle, NumSamples)
for I := 0 to NumSamples-1 by 1
    get_sample_class_mlp (MLPHandle, I, Data, Target)
    classify_class_mlp (MLPHandle, Data, 1, Class, Confidence)
    Result := gen_tuple_const(NOut,0)
    Result[Class] := 1
    Diffs := Target-Result
    if (sum(fabs(Diffs)) > 0)
        * Sample has been classified incorrectly
    endif
endfor
clear_class_mlp (MLPHandle)

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

get_sample_class_mlp is reentrant and processed without parallelization.

add_sample_class_mlp, read_samples_class_mlp, get_sample_num_class_mlp

classify_class_mlp, evaluate_class_mlp

create_class_mlp

Foundation