off-key
A platform for online anomaly detection, connecting sensor data, conformal inference, and monitoring interfaces. Developed within BiFlex-Industrie for charging infrastructure, with broader applications in mind.
Machine learning & statistical inference
Oliver Hennhöfer
I research uncertainty in machine learning and develop software that puts statistical methods into practice.
I’m a research associate and lecturer at Karlsruhe University of Applied Sciences, working on conformal inference, anomaly detection, and sequential testing. My background spans geospatial analysis, machine learning in public transport, and research software development.
Research software
nonconform
Conformal anomaly detection in Python: score calibration, false discovery rate control, and sequential change monitoring. Integrates with scikit-learn, PyOD, and custom detectors.
online-fdr
Sequential hypothesis testing with p-values and e-values. Python implementations of online false discovery rate control.
aberrant
Online anomaly detection for streaming data. Python methods and tooling for identifying unusual observations as a stream unfolds.
talk-tag
Publications
- Between Resolution Collapse and Variance Inflation: Weighted Conformal Anomaly Detection in Low-Data Regimes
- TalkTag: Fine-Grained Morphosyntactic Error Annotation for Transcribed Speech
- Conformal Anomaly Detection in Python: Moving Beyond Heuristic Thresholds with nonconform
- Interdisciplinary Harmonies: A Story-Driven Course on AI and Music to Increase Interest in Computer Science
- Leave-One-Out-, Bootstrap- and Cross-Conformal Anomaly Detectors
- Multidimensional Exploratory Spatial Data Analysis