Computer implementation methods, computer-controlled systems, computer programs (neural network systems for abstract justification) that solve cognitive tasks, including learning abstract properties.

摘要

PROBLEM TO BE SOLVED: To implement a computer implementation method, a system, and a computer program product for solving a cognitive task, including learning abstract characteristics. One embodiment may include accessing a dataset that characterizes an abstract property. The accessed dataset can then be input to the first neural network to generate the first embedding. The generated first embedded pair can be formed, which corresponds to the pair of datasets. The data corresponding to the formed pair is then input to the second neural network, which can be performed to generate a second embedding. The latter can capture the relevant characteristics of a pair of datasets. A third neural network is subsequently executed based on the generated second embedding to obtain the output value. To solve the cognitive task, one or more abstract properties of the dataset are learned based on the acquired output values. [Selection diagram] Fig. 1