ABNORMALITY DETECTION DEVICE FOR MACHINE-LEARNING DEVICE, PERSON DETECTION SYSTEM, AND ABNORMALITY DETECTION METHOD FOR MACHINE-LEARNING DEVICE

摘要

According to the present invention, after learning of a machine-learning device (100) is completed, an abnormality detection range setting unit (530): acquires, from the machine-learning device (100), a determination result obtained by inputting input data to the machine-learning device (100); selects, from among a plurality of pieces of input data, a plurality of pieces of input data determined, by the machine-learning device (100), to belong to a first category; specifies a minimum value and a maximum value for each dimension of the plurality of pieces of selected input data; and sets, as an abnormality detection range, an N-dimensional area determined by the minimum value and the maximum value of each dimension. During operating of the machine-learning device (100), an abnormality determination unit (520) determines the machine-learning device (100) to be abnormal when the input data to the machine-learning device (100) is determined to fall within the abnormality detection range and the input data to the machine-learning device (100) is determined to belong to a second category.