Fecha de publicación: 11/07/2019 Fuente: WIPO "abeja" An objective of the present invention is to enable the derivation, from a learned neural network (NN), of a plurality of smaller NNs of varying scale. Provided is a data processing system comprising: a neural network (NN) comprising multiplexed constituent elements such as multiple layers, dimensions, or channels; an optimizer for, in an NN learning process, repeating a learning session for correcting a setting of a parameter comprised in the NN; and a chopper for, in the NN learning process, reducing the multiplicity of the multiplexed constituent elements of the NN. The chopper reduces the multiplicity of the multiplexed constituent elements variably on the basis of a sequence assigned to the multiplexed constituent elements.