Information processing apparatus and information processing method

Abstract

The purpose of the present invention is to obtain a high-precision processing result even if a sample in a target domain does not sufficiently correspond to a sample in a source domain. The invention relates to an information processing apparatus and an information processing method. An information processing device includes: a feature extraction unit that extracts features from first and second samples included in a source domain and from the first sample included in a target domain; a suspected sample generation unit that generates a second-class suspected sample of the target domain on the basis of a distribution of the first-class samples included in the target domain in the feature space of the features extracted by the feature extraction unit; and a data conversion unit that performs data conversion by mechanical learning such that the distribution of the first-class samples and the second-class samples included in the source domain in the feature space approaches the distribution of the first-class samples and the second-class suspected samples included in the target domain.

Description

technical field [0001] The present invention relates to an information processing device and an information processing method, and in particular, to a technology for making a learning model learn mechanically. Background technique [0002] In taught machine learning, a teaching data set including teaching samples (specimens) and teaching labels is used, and the learning model is machine-learned by reflecting the relationship between the teaching samples and labels to the learning model. By applying the learning completion model obtained by such machine learning to unknown samples without labels in the inference stage, for example, desired processing results such as image recognition and classification are obtained. [0003] If a learning completion model learned using a certain teaching data set is directly applied to a task targeting a data set of a different domain from the teaching data set, the accuracy of inference will be reduced. Here, the domain means the kind, rang...

AI Technical Summary

Problems solved by technology

[0003] If a learned model learned using a certain teaching dataset is directly applied to a task targeting a dataset of a different domain than the teaching dataset, it will lead to a reduction in the accuracy of the inference

Images