Raytheon Company Patent for ML Model Drift Detection

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted Raytheon Company a patent for a method of detecting machine learning (ML) model drift using a modified generative adversarial network (GAN). The patent details a system that monitors the output of a deployed ML model to identify drift and determine if re-deployment is necessary.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has issued patent US12585918B2 to Raytheon Company for an invention related to machine learning (ML) model drift detection. The patent describes a method that utilizes a modified generative adversarial network (GAN) architecture to monitor the output of a deployed ML model. Specifically, it involves recording the output of a hidden layer, determining a metric from this output, and using this metric to assess whether the model is suffering from drift, potentially triggering a re-deployment.

This patent grant is primarily an intellectual property matter and does not impose direct regulatory obligations on companies. However, it signifies innovation in the field of AI and ML model monitoring. Companies developing or deploying ML models, particularly those in sectors where model accuracy is critical, may find the patented techniques relevant for their internal development or as a basis for licensing discussions. The filing date for this patent was October 21, 2022, with the grant date being March 24, 2026.

Source document (simplified)

← USPTO Patent Grants

ML model drift detection using modified GAN

Grant US12585918B2 Kind: B2 Mar 24, 2026

Assignee

Raytheon Company

Inventors

Nicole M. Hatten

Abstract

Discussed herein are devices, systems, and methods for machine learning (ML) model drift detection. A method can include receiving machine learning (ML) data defining a number of layers of neurons, a number of neurons per each layer, and weights for each neuron of a deployed ML model, operating the deployed ML model in a modified generative adversarial network (GAN) architecture, while operating the deployed ML model, recording output of a hidden layer of the deployed ML model, determining a metric of the output, and re-deploying the deployed ML model and monitoring whether the re-deployed ML model is suffering from ML model drift based on the metric.