Machine learning

Description

G06N20/00 is the main group for machine learning, i.e. computing arrangements in which behaviour is derived from training data rather than fixed programming. It spans supervised, unsupervised and reinforcement learning workflows, including model training, validation and inference infrastructure. Kernel methods such as support vector machines are classified in G06N20/10 and ensemble techniques such as boosting or random forests in G06N20/20. Neural-network models as such belong to G06N3/00, and the application of learned models to pattern recognition tasks is covered by G06F18/00.

Filing statistics

IPC G06N 20/00 belongs to the parent subclass G06N, which recorded 613,242 patent applications worldwide between 2013 and 2023. Annual filings grew from 3,903 in 2013 to 139,015 in 2023 (+3462%). The most active applicants in class G06 are SAMSUNG ELECTRONICS COMPANY (66,669 applications), IBM (INTERNATIONAL BUSINESS MACHINES CORPORATION) (62,313 applications) and MICROSOFT TECHNOLOGY LICENSING (41,918 applications).

Source: EPO PATSTAT, worldwide filings