Microsoft patents method to stop training image upscaling models early
Microsoft describes a method for training software that makes low-resolution images look sharper. The system checks how quickly the software improves after each round of practice and stops the process when the improvements slow down to a specific level.
In the filing’s own words
Systems and methods are provided for dynamically modifying the training of image processing models. In some instances, a set of training data is obtained that includes pairings of low-resolution and correlated high-resolution images. The training data is separated into a training group and a validation group. During a training cycle, a image processing model is applied to the training group over multiple epochs. After each epoch, the model is applied to the validation data to calculate a loss value. A rate of change in the loss value between the different epochs is determined, and a predetermined rate of change threshold is identified. The training cycle is terminated when the rate of change in the loss value meets or is less than the predetermined threshold.
What exactly they claim (claim 1)
1. A method for training an image processing model that is configured to transform images from a first resolution to a second resolution, the second resolution comprising a higher resolution than the first resolution, comprising: obtaining a set of training data comprising a plurality of correlated image pairings, each correlated image pairing including a first image at a first resolution and a correlated second image at a second resolution that is higher than the first resolution; training the image processing model with the training data during a training cycle that includes a plurality of epochs of the image processing model processing the training data; after each epoch, applying the image processing model to validation data associated with the training data to calculate a loss value; determining a rate of change in the calculated loss value between different epochs; identifying a predetermined rate of change threshold associated with terminating the training cycle; and terminating the training cycle when the rate of change in the calculated loss value meets or is less than the predetermined rate of change threshold.
Facts
- Company
- Microsoft (Microsoft Technology Licensing, LLC)
- Official title
- TRAINING PROCESSES FOR MACHINE LEARNING SUPER-RESOLUTION MODELS
- Publication number
- US20260359933A1
- Application number
- 19743155
- Filed
- Jul 15, 2026
- Published
- Nov 12, 2026
- Inventors
- 2
- Claims
- 20
Classified under
G06T 3/4053 Physics & computingG06T 5/50 Physics & computingG06T 5/60 Physics & computingG06T 2207/20016 Physics & computingG06T 2207/20081 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.
