Rondo uncovers repeated patterns in streaming time series data
Rondo: Unsupervised Discovery of Recurring Temporal Structure
Machine LearningArtificial Intelligence
Summary
Many types of data that change over time, like sounds or sensor readings, have smaller repeated parts that form bigger repeated sequences. The authors present Rondo, a way to automatically find these repeating parts and how they combine, even when new data keeps coming in. Unlike older methods, Rondo recognizes patterns at different time lengths as connected and updates what it knows as it sees more data. This helps machines understand long streams of time-based data better without needing lots of labeled examples. The authors tested Rondo on different kinds of data and it worked better, especially when data was limited or continuously arriving.
What this means in practice
- •For data engineers: Identify and update repeated temporal patterns in large, unlabeled streaming data to improve data interpretation pipelines.
- •For speech recognition developers: Build adaptive models that discover and refine words and phrases from continuous audio streams without manual labeling.
Authors
Yingtian Shi, Ankith Chandra, Thomas Plötz
Abstract
Many real-world time series data exhibit structural properties at multiple scales, from short, recurring units to complex sequences composed of these units. Unsupervised discovery of both these components and structure enables the design of intelligent systems that help interpret temporal data, thereby limiting the amount of costly human annotations required. Existing modeling approaches typically overlook the hierarchical structure inherent to many time series, treating recurring patterns at different temporal scales as independent structures. Moreover, most assume access to the complete data sequence and treat discovery as a static process, limiting their ability to evolve as new observations arrive. We introduce Rondo, an unsupervised approach for modeling recurring hierarchical structure in continuous temporal streams. By explicitly constructing vocabularies of reusable units and their recurring compositions, Rondo captures structure shared across complex temporal patterns while refining and expanding its discoveries as the stream evolves. Evaluations on temporal sequences spanning diverse domains and data modalities show that Rondo outperforms existing unsupervised recurrence-discovery baselines, with particularly pronounced advantages in limited-data and continual-stream settings. These capabilities provide a stronger foundation for recurring-pattern discovery, scalable behavior understanding, and adaptive intelligent systems operating on long, unlabeled temporal streams.