Machine Learning

Machine learning is the branch of artificial intelligence in which a system learns from examples instead of following rules a programmer wrote out. NIST's glossary defines it as "The development and use of computer systems that adapt and learn from data with the goal of improving accuracy" (NIST SP 800-55v1, in the CSRC glossary). GAO uses "model" for "the result of an algorithm 'trained' on a set of data," and calls training "the iterative process of feeding data (called training data) through an optimization process to improve model performance" (GAO-24-106946, 2024).

Main types

NIST's adversarial machine learning glossary (NIST AI 100-2e2025) defines the common forms:

• Supervised learning, in which "a model learns to predict explicit (often human-generated) labels or output values for data."
• Unsupervised learning, in which "a model learns based on patterns in unlabeled data, such as learning a function to cluster or group data points."
• Semi-supervised learning, in which "a small number of training samples are labeled, while the majority are unlabeled."
• Reinforcement learning, in which "a model learns to optimize its behavior according to a reward function by interacting with and receiving feedback from an environment."

A basic task is classification: "The task of predicting which of a set of discrete categories an input belongs to."

A records example

The National Archives described machine learning for sorting federal records in its 2014 Automated Electronic Records Management Report. In "autocategorization with machine learning, an expert trains the system to recognize records that fit in each retention category based on categorization of a training set and iterative reviews of additional machine-coded documents." The expert "never strictly defines the characteristics of the category as would be necessary to write an executable rule." NARA cautioned that "the work by experts required to train the systems is significant." Rule-based alternatives are covered under Automation.

Limits

A model is built from its data, so the data sets its ceiling. NIST warns that training datasets "may become detached from their original and intended context or may become stale or outdated relative to deployment context," and that AI systems "may require more frequent maintenance and triggers for conducting corrective maintenance due to data, model, or concept drift" (NIST AI 100-1, Appendix B). Machine learning built on many-layered neural networks is covered under Deep Learning.

Why it matters for land and mining claim records

A model trained on one set of documents learns that set. Forms, typefaces, and filing conventions differ between offices and decades, which is the drift NIST describes. The practical check is to compare machine results against a sample of source documents from each new batch.