Device and computer-implemented process for data-efficient active machine learning

摘要

Device and computer-implemented process for active machine learning, providing recorded data (206),where a model is trained to classify the data according to the data recorded (208),where values of an acquisition function of the non-recorded data are determined for non-recorded data (210),and where non-recorded data are used to acquire non-recorded data for active mechanical learning (212),whose values for the acquisition function meet a criterion whereby auto-correlation is determined by a feature representation for a sample to be evaluated from the non-recorded data, in particular from at least one layer of the model, where the value for the acquisition function of this model is determined on the basis of a square mean over autocorrelation, in particular in at least one dimension.