11-08, 15:20–16:00 (US/Eastern), Central Park East
PyNomaly's Local Outlier Probability (LoOP) implementation provides interpretable and reliable detection capabilities for classical anomaly types in Python for both static and streaming data.
Anomaly detection is an important capability across many industries and applications, including (but not limited to) finance, aerospace, and research applications. When applying anomaly detection approaches, the interpretability of an anomaly score is an important consideration in ensuring implementation success. In this talk, attendees will learn about PyNomaly and it's implementation of Local Outlier Probabilities (LoOP), including the algorithm's convenient interpretability that is free of the underlying statistical distribution of the data, it's potential applications as well as it's limitations for detecting certain anomaly types which require more advanced methods. Attendees will leave understanding how to leverage PyNomaly for their anomaly detection and data analysis projects.
Slides: https://docs.google.com/presentation/d/1YgmbNOuAWlLIQKvEcnsGmv5-lzSz7CTZdkCFQahndW4/edit?usp=sharing
Previous knowledge expected
I love the process of building impactful products and services, from zero to one. Currently building at Infactory: https://www.infactory.ai/
At Terran Orbital, I established a strong internal artificial intelligence capability, building and leading a team of data scientists and machine learning engineers, developing cloud-native, event-driven scalable platforms for remote sensing geospatial analytics and supporting cross-functional internal process automation sprints. The team reduced hardware related production-related commissioning times by over 85%, enabling scalable production of that externally sold component.
At the NASA Jet Propulsion Laboratory ("JPL", operated by the California Institute of Technology, "CalTech"), I served as the Principal Investigator to a multi-year alarm analytics effort, co-organized a monthly meetup of the Lab's open source developers (the Open Developer Meetup), and lead innovative applied machine and deep learning research and development efforts. I released the open-source PyNomaly software during my time there and continue to maintain the software - a core library in anomaly detection.
Ekin Tiras is a Senior Software Developer at SAP, specialising in observability and designing automated telemetry analysis systems with machine learning. With a focus on improving observability solutions, Ekin designs and implements models to drive more insightful analytics. Previously, he worked at a major German public broadcaster, where he contributed to a platform that leverages machine learning to extract and categorize metadata from video files. His professional work and personal interest in anomaly detection have also led him to become a maintainer of the open-source library PyNomaly in his spare time.