Preliminary Results for Using Uncertainty and Out-of-distribution Detection to Identify Unreliable Predictions.

Technical Report · Wed Nov 23 04:00:00 EST 2022

DOI:https://doi.org/10.2172/1899654· OSTI ID:1899654

Doak, Justin E. ^[1]; Darling, Michael C. ^[1]

---

1. Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)

As machine learning (ML) models are deployed into an ever-diversifying set of application spaces, ranging from self-driving cars to cybersecurity to climate modeling, the need to carefully evaluate model credibility becomes increasingly important. Uncertainty quantification (UQ) provides important information about the ability of a learned model to make sound predictions, often with respect to individual test cases. However, most UQ methods for ML are themselves data-driven and therefore susceptible to the same knowledge gaps as the models themselves. Specifically, UQ helps to identify points near decision boundaries where the models fit the data poorly, yet predictions can score as certain for points that are under-represented by the training data and thus out-of-distribution (OOD). One method for evaluating the quality of both ML models and their associated uncertainty estimates is out-of-distribution detection (OODD). We combine OODD with UQ to provide insights into the reliability of the individual predictions made by an ML model.

• Research Organization: Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States); Sandia National Laboratories (SNL-CA), Livermore, CA (United States)
• Sponsoring Organization: USDOE National Nuclear Security Administration (NNSA)
• DOE Contract Number: NA0003525
• OSTI ID: 1899654
• Report Number(s): SAND2022-16321; 712192
• Country of Publication: United States
• Language: English

Related Subjects

97 MATHEMATICS AND COMPUTING