Companies with the most machine learning and AI patents in Japan 2020

* A patent family is defined as a group of patent documents all of which have the same set of priorities. I.e., it is a group of patent documents that protect the same invention across authorities.

The owner of a patent includes all patents owned by a company and its subsidiaries (by at least 50%) based on comprehensive research on corporate structures, M&A, Spin-offs, company names changes, patent transactions. PatentSight also tracks former names of companies and subsidiaries, ongoing mergers and acquisitions, acquisitions of the past, reassignments, companies that have the same applicant, and transliterations by patent offices.

Technology field definition by Swiss Federal Institute of Intellectual Property – ip-search: "This field covers machine learning as an ubiquitous technology employed in a variety of technical applications, such as image, text or speech analysis in areas from digital data analysis, predictive maintenance to autonomous driving to name only few. In this technology field, machine learning is defined as signal and data processing algorithms, implementing devices and systems which involve schemes based on given observations and outcomes. Distinction to calibration and parametrization, as well as rule-based systems is made in that documents in this field explicitly aim at optimizing generalization abilities of the proposed invention. Thus, the field contains inventions solving classification, clustering and regression in high dimensional data spaces as well as patents describing artificial intelligence of any form. Modelling paradigms include biological inspired approaches, e.g. neural networks or genetic algorithms, kernelized methods such as support vector machines, reinforced or deep learning."