USPTO Patent US12585964B2: Exhaustive learning techniques for machine learning algorithms

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted patent US12585964B2 to PayPal, Inc. for "Exhaustive learning techniques for machine learning algorithms." This patent details iterative machine learning operations, including training models and updating training datasets.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has issued patent US12585964B2 to PayPal, Inc. The patent covers "Exhaustive learning techniques for machine learning algorithms," specifically detailing methods for iterative machine learning operations. These methods involve training a machine learning model, identifying and removing subsets of data samples that meet certain criteria to update the training dataset, and repeating this process to achieve a final trained model.

This patent grant is primarily an intellectual property matter and does not impose direct compliance obligations on regulated entities. However, it signifies innovation in AI and machine learning techniques, which may influence future technological developments and competitive landscapes within the technology sector. Companies operating in AI and machine learning should be aware of this patent as it pertains to their intellectual property strategy and potential licensing considerations.

Source document (simplified)

← USPTO Patent Grants

Exhaustive learning techniques for machine learning algorithms

Grant US12585964B2 Kind: B2 Mar 24, 2026

Assignee

PayPal, Inc.

Inventors

Zeding Li

Abstract

Techniques are disclosed relating to exhaustive learning techniques for machine learning algorithms. In various embodiments, the disclosed techniques include performing an iterative machine learning operation that includes training a first version of a machine learning model (e.g., a decision tree model) based on a current version of a training dataset, where the first version of the machine learning model includes a plurality of decision branches, identifying a first subset of data samples that satisfy evaluation criteria included in a first one of the plurality of decision branches, and removing this first subset of data samples to generate an updated version of the training dataset. In various embodiments, the disclosed techniques include repeating the iterative machine learning operation using the updated version of the training dataset to produce a final trained version of the machine learning model.