Processing of learning data sets including noisy labels for classifiers

A method for processing of learning data sets for a classifier, the learning data sets each encompassing measured data, which were obtained by a physical measuring process, and/or by a partial or complete simulation of a physical measuring process, and/or by a partial or complete simulation of a technical system observable using a physical measuring process, as learning input variable values, the learning data sets additionally each encompassing learning output variable values to which the classifier is to nominally map the learning input variable values, the method comprising the following steps:

• processing the learning input variable values of at least one of the learning data sets multiple times in a non-congruent manner by the classifier trained up to an epoch E2 so that the learning input variable values are mapped to different output variable values;
• ascertaining a measure for the uncertainty of the output variable values from the deviations of the output variable values;
• in response to the uncertainty meeting a predefined criterion, ascertaining at least one updated learning output variable value for the learning data set from one or multiple further output variable value(s) to which the classifier maps the learning input variable values after a reset to an earlier training level with epoch E1<E2,
• wherein the learning input variable values are mapped by multiple modifications of the classifier, which differ from one another to such an extent that they are not congruently merged into one another with advancing training, to different output variable values, and
• wherein a separate uncertainty is ascertained for each output variable value ascertained by each of the modifications, only an established number of the modifications which generated the output variable values having lowest uncertainties contributing to the ascertainment of the updated learning output variable value.

A method for training a classifier based on learning data sets, each of the learning data sets encompassing measured data, which were obtained by a physical measuring process, and/or by a partial or complete simulation of a physical measuring process, and/or by a partial or complete simulation of a technical system observable using a physical measuring process, as learning input variable values, each of the learning data sets additionally encompassing learning output variable values to which the classifier is to nominally map the learning input variable values, the method comprising the following steps:

• training a preclassifier to distinguish learning data sets whose respective learning input variable values are assigned correct learning output variable values from learning data sets whose respective learning input variable values are assigned noisy or incorrect learning output variable values;
• classifying the learning data sets using the trained preclassifier, and adding those of the learning data sets, whose respective learning input variable values are assigned correct learning output variable values, to a set of preferred learning data sets;
• providing at least one learning data set of the learning data sets, whose learning input variable values are assigned noisy or incorrect learning output variable values with at least one updated learning output variable value;
• adding the learning data set with the at least one updated learning output variable value to the set of preferred learning data sets;
• and
• training the classifier using the learning data sets in the set of preferred learning data sets.

A method, comprising the following steps:

• training at least one classifier based on processed learning data sets, the learning data sets each encompassing measured data, which were obtained by a physical measuring process, and/or by a partial or complete simulation of a physical measuring process, and/or by a partial or complete simulation of a technical system observable using a physical measuring process, as learning input variable values, the learning data sets additionally each encompassing learning output variable values to which the classifier is to nominally map the learning input variable values, the learning data sets being processed by:
• processing the learning input variable values of at least one of the learning data sets multiple times in a non-congruent manner by the classifier trained up to an epoch E2 so that the learning input variable values are mapped to different output variable values,
• ascertaining a measure for the uncertainty of the output variable values from the deviations of the output variable values,
• in response to the uncertainty meeting a predefined criterion, ascertaining at least one updated learning output variable value for the learning data set from one or multiple further output variable value(s) to which the classifier maps the learning input variable values after a reset to an earlier training level with epoch E1<E2,
• wherein the learning input variable values are mapped by multiple modifications of the classifier, which differ from one another to such an extent that they are not congruently merged into one another with advancing training, to different output variable values, and
• wherein a separate uncertainty is ascertained for each output variable value ascertained by each of the modifications, only an established number of the modifications which generated the output variable values having lowest uncertainties contributing to the ascertainment of the updated learning output variable value;
• operating the classifier by supplying input variable values to it, the input variable values encompassing measured data which were obtained by a physical measuring process, and/or by a partial or complete simulation of a physical measuring process, and/or by a partial or complete simulation of a technical system observable using a physical measuring process;
• forming an activation signal as a function of output variable values supplied by the classifier;
• and
• activating, using the activation signal, a vehicle and/or a system for the quality control of series-manufactured products, and/or a system for medical imaging.

A non-transitory machine-readable data carrier on which is stored a computer program for processing of learning data sets for a classifier, the learning data sets each encompassing measured data, which were obtained by a physical measuring process, and/or by a partial or complete simulation of a physical measuring process, and/or by a partial or complete simulation of a technical system observable using a physical measuring process, as learning input variable values, the learning data sets additionally each encompassing learning output variable values to which the classifier is to nominally map the learning input variable values, the computer program, when executed by a computer, causing the computer to perform the following steps:

• processing the learning input variable values of at least one of the learning data sets multiple times in a non-congruent manner by the classifier trained up to an epoch E2 so that the learning input variable values are mapped to different output variable values;
• ascertaining a measure for the uncertainty of the output variable values from the deviations of the output variable values;
• in response to the uncertainty meeting a predefined criterion, ascertaining at least one updated learning output variable value for the learning data set from one or multiple further output variable value(s) to which the classifier maps the learning input variable values after a reset to an earlier training level with epoch E1<E2,
• wherein the learning input variable values are mapped by multiple modifications of the classifier, which differ from one another to such an extent that they are not congruently merged into one another with advancing training, to different output variable values, and
• wherein a separate uncertainty is ascertained for each output variable value ascertained by each of the modifications, only an established number of the modifications which generated the output variable values having lowest uncertainties contributing to the ascertainment of the updated learning output variable value.

A computer configured to process learning data sets for a classifier, the learning data sets each encompassing measured data, which were obtained by a physical measuring process, and/or by a partial or complete simulation of a physical measuring process, and/or by a partial or complete simulation of a technical system observable using a physical measuring process, as learning input variable values, the learning data sets additionally each encompassing learning output variable values to which the classifier is to nominally map the learning input variable values, the computer configured to:

• process the learning input variable values of at least one of the learning data sets multiple times in a non-congruent manner by the classifier trained up to an epoch E2 so that the learning input variable values are mapped to different output variable values;
• ascertain a measure for the uncertainty of the output variable values from the deviations of the output variable values;
• in response to the uncertainty meeting a predefined criterion, ascertain at least one updated learning output variable value for the learning data set from one or multiple further output variable value(s) to which the classifier maps the learning input variable values after a reset to an earlier training level with epoch E1<E2,
• wherein the learning input variable values are mapped by multiple modifications of the classifier, which differ from one another to such an extent that they are not congruently merged into one another with advancing training, to different output variable values, and
• wherein a separate uncertainty is ascertained for each output variable value ascertained by each of the modifications, only an established number of the modifications which generated the output variable values having lowest uncertainties contributing to the ascertainment of the updated learning output variable value.