Brent Heeringa

Papers

Total Citations

H-Index

About

Brent Heeringa is a computer scientist whose research spans unsupervised machine learning, hierarchical agent control, and natural language processing. His most influential work focuses on the unsupervised segmentation of categorical time series into meaningful episodes, where he developed the VOTING-EXPERTS algorithm. This innovative approach collects n-gram statistics—frequency and boundary entropy—then employs two expert methods within a sliding window to determine segmentation boundaries, enabling autonomous discovery of patterns in sequential data without labeled training examples. His 2003 paper on this topic has garnered 30 citations, with a closely related 2002 publication adding 19 more. Heeringa also made significant contributions to multi-agent systems through the Hierarchical Agent Control (HAC) architecture, which provides a general toolkit for specifying agent behavior with action abstraction, resource management, and sensor integration—particularly suited for controlling large numbers of agents in dynamic environments. This work has accumulated 29 citations. Additionally, Heeringa explored learning stochastic context-free grammars for natural language syntax, addressing the memory and computational challenges that make traditional algorithms impractical for embedded agents. His research demonstrates a consistent focus on developing efficient, unsupervised methods for pattern discovery and autonomous control in resource-constrained settings.