Microsoft patents method to stop AI training early using data patterns
Microsoft describes a technique that measures the complexity of data using a mathematical concept called fractal dimension to decide when a machine learning model has learned enough. By monitoring this metric, the system can stop training early if the data patterns stabilize, potentially saving significant time without reducing accuracy.
In the filing’s own words
Some embodiments select a machine learning model training duration based at least in part on a fractal dimension calculated for a training data dataset. Model training durations are based on one or more characteristics of the data, such as a fractal dimension, a data distribution, or a spike count. Default long training durations are sometimes replaced by shorter durations without any loss of model accuracy. For instance, the time-to-detect for a model-based intrusion detection system is shortened by days in some circumstances. Model training is performed per a profile which specifies particular resources or particular entities, or both. Realistic test data is generated on demand. Test data generation allows the trained model to be exercised for demonstrations, or for scheduled confirmations of effective monitoring by a model-based security tool, without thereby altering the model's training.
What exactly they claim (claim 1)
1. A computer-implemented method comprising: obtaining a time series dataset having a dataset duration; calculating a fractal dimension of the time series dataset based on training data accumulated over the dataset duration; training a machine learning model using the training data; monitoring changes in the fractal dimension as additional training data is accumulated; determining that the fractal dimension satisfies a convergence condition based on the monitored changes; and terminating the training of the machine learning model based on the convergence condition.
Facts
- Company
- Microsoft (Microsoft Technology Licensing, LLC)
- Official title
- MACHINE LEARNING TRAINING DURATION CONTROL
- Publication number
- US20260289413A1
- Application number
- 19682853
- Filed
- May 20, 2026
- Published
- Sep 24, 2026
- Inventors
- 3
- Claims
- 20
Classified under
G06N 20/00 Physics & computingG06N 3/08 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.