Microsoft patents system that adjusts AI training steps based on performance
Microsoft describes a system for training artificial intelligence models across multiple separate data locations, known as silos. The system automatically adds or removes training steps depending on whether the model meets specific performance goals.
In the filing’s own words
A data processing system implements receiving, in a first silo, configuration parameters for performing federated training of the machine learning model using data stored in the plurality of second silos, generating a federated learning pipeline for automatically executing a plurality of tasks associated with the federated learning of the machine learning model based on the configuration parameters, causing each respective silo of the plurality of second silos to train the local instance of the machine learning model according to the federated learning pipeline, obtaining, at the first silo, the learnable parameters information associated with the local instance of the machine learning model of each silo of the plurality of second silos, and aggregating the learnable parameters information obtained from the plurality of second silos to obtain a primary instance of the machine learning model that has been fine-tuned according to the learnable parameters information.
What exactly they claim (claim 1)
1. A data processing system comprising: a processor; and a machine-readable medium storing executable instructions that, when executed, cause the processor to perform operations comprising: receiving, in a first silo, configuration parameters for performing federated learning of a machine learning model using data stored in a plurality of second silos; generating a federated learning pipeline, the federated learning pipeline further including a plurality of training iteration stages, for automatically executing a plurality of tasks associated with the federated learning of the machine learning model based on the configuration parameters; causing a respective silo of the plurality of second silos to train a local instance of the machine learning model according to the federated learning pipeline; obtaining, at the first silo, learnable parameters information associated with the local instance of the machine learning model of each silo of the plurality of second silos; aggregating the learnable parameters information obtained from the plurality of second silos to obtain a primary instance of the machine learning model that has been fine-tuned according to the learnable parameters information, performing a determination of whether the performance of the primary instance of the machine learning model satisfies one or more performance criteria based on outputs of the machine learning model and reference output data; and responsive to the determination, dynamically altering a flow of the federated learning pipeline, including automatically removing one or more of the plurality of training iteration stages from the federated learning pipeline when the one or more performance criteria is satisfied or automatically adding one or more of the plurality of training iteration stages to the federated learning pipeline when the one or more performance criteria is not satisfied.
Facts
- Company
- Microsoft (Microsoft Technology Licensing, LLC)
- Official title
- Federated Learning Solutions for Machine Learning
- Publication number
- US20260278483A1
- Application number
- 19675610
- Filed
- May 13, 2026
- Published
- Sep 17, 2026
- Inventors
- 3
- Claims
- 20
Classified under
G06N 20/20 Physics & computingG06F 11/3495 Physics & computingG06F 21/6245 Physics & computingMore from Microsoft in AI & software
- Microsoft patents method to check AI answers against verified facts Oct 20, 2026
- Microsoft patents AI that presents PowerPoint slides for you Sep 3, 2026
- Microsoft patents method to check if LLM answers are based on real data Sep 10, 2026
- Microsoft patents system to blend multiple AI agents into one chat Oct 1, 2026
- Microsoft patents system that answers questions using your screen history Oct 15, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.