USPTO Grants Patent for Machine Learning Model Dropout Explanation

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted a patent (US12585996B2) to Fair Isaac Corporation for a machine learning model dropout explanation system. This patent covers methods for improving computer-implemented machine learning models by evaluating the impact of removing key features to explain model outputs.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has granted patent US12585996B2 to Fair Isaac Corporation for "Explanatory dropout for machine learning models." This patent details systems and methods for enhancing machine learning models by using "on-manifold/on-distribution evaluation of dropout of key features" to explain model outputs. The patented technology allows for the removal of influence from specific features to provide clearer explanations of model predictions, offering an advantage over traditional explanation methods.

This patent grant is primarily an intellectual property development and does not impose direct regulatory obligations on companies. However, it signifies a new development in AI explainability technology. Companies developing or utilizing machine learning models, particularly in areas where model transparency and explainability are critical (e.g., finance, healthcare), should be aware of this patented technology. While not a compliance mandate, understanding existing patents in AI can inform R&D strategies and potential licensing considerations.

Source document (simplified)

← USPTO Patent Grants

Explanatory dropout for machine learning models

Grant US12585996B2 Kind: B2 Mar 24, 2026

Assignee

Fair Isaac Corporation

Inventors

Matthew Kennel, Scott Zoldi

Abstract

Explanatory dropout systems and methods for improving a computer implemented machine learning model are provided using on-manifold/on-distribution evaluation of dropout of key features to explain model outputs. The machine learning model is trained using a plurality of input examples, including input records with explicit dropout operators applied effectuating the removal of influence of features associated with an explanation reason class. One or more dropout operators may be stochastically applied to one or more input examples. The procedure includes on-manifold/on-distribution evaluation of the machine learning model under conditions of absence or presence of the one or more dropout operators for reliable calculation of numerical statistics associated with reason classes to yield model explanations. The training and evaluation procedures present advantages over traditional off-manifold or off-distribution perturbative explanation procedures.