AI & Functional Safety: Out-of-Distribution Detection as a Safety Mechanism
- Date
- 2027-05-31
- Location
- Online
- Host
- Zach L
About this event
A live 30-minute expert session on Out-of-Distribution Detection as a Safety Mechanism (AI / ML — AI & Machine Learning Safety). What We'll Cover: What OOD detection is and why distributional shift is the dominant failure mode for deployed ML Core methods step by step: softmax confidence, energy scores, autoencoder reconstruction, and ensembles — and how to pick How an OOD monitor maps to a safety mechanism claim in ISO 26262/ISO 21448 terms and the detection-rate evidence it produces Common mistakes: tuning thresholds on in-distribution data only and ignoring detector latency in the fault-reaction time budget The false-negative/false-positive characterization and ODD-shift test evidence an auditor expects Related topics: out of distribution detection · ood detection · distributional shift