FMfSS2019: Formal Methods for Statistical Software Historic Inns of Annapolis Annapolis, MD, United States, May 1, 2019 |
Conference website | https://samate.nist.gov/FMSwVRodeo/FMfSS2019.html |
Submission link | https://easychair.org/conferences/?conf=fmfss2019 |
Submission deadline | March 2, 2019 |
Statistical software and related data analysis pose different challenges than typical software. A simplistic pseudorandom number generator may offer sufficient randomness for a video game, but be inadequate for valid Monte Carlo simulation or other stochastic algorithms. Likewise, erroneous implementations of mathematical functions may generate results that look correct, but which fail in unpredictable ways—examples of which include both errors in floating point division implementations and bugs in differential privacy implementations that leak confidential data into supposedly “privatized” outputs. Such errors are typically missed by traditional testing.
Formal methods can complement testing to gain greater assurance that critical portions of programs are correct and that results are valid. Although there is a vast expanse of formal methods, tools, and techniques, they are rarely applied to such software. This workshop asks, what formal methods can mitigate, detect, correct, or preclude flaws in statistical software or errors of incorrect use?
Submission Guidelines
We call for position statements from one to three paragraph long. Based in part on position statements, the program committee will invite presentations at the workshop. Position statements will be published if agreed to by both the author and the program committee.
List of Topics
- types of errors, bugs, and failures in statistical software or its use
- formal methods that can assure statistical software or its results
- correct-by-construction approaches
- lists or definitions of important properties, such as convergence and differential privacy
- gaps in current capabilities and specific research needed
- higher-level, non-procedural languages or tools so a statistician can specify the desired analysis instead of how to achieve it
- approaches to gain assurance in statistical software and its use
Co-Chairs
- Paul E. Black (NIST)
- Simson L. Garfinkel (U.S. Census Bureau)
Program Committee
- Jim Alves-Foss (U Idaho)
- Rance Cleaveland (NSF)
- Alan Heckert (NIST)
- Martin Klein (U.S. Census Bureau)
- Rick Kuhn (NIST)
- Stephen Magill (Galois)
- Edward "Ned" Porter (U.S. Census Bureau)
- John Henry Scott (NIST)
- Ray Richards (DARPA)
- Eric Smith (Kestrel Institute)
- Matthew Wilding (Collins Aerospace)
- Huan Xu (U Maryland)
Venue
The workshop will be at Historical Inns in Annapolis, Maryland, USA.
Contact
All questions about submissions should be emailed to Paul E. Black paul.black@nist.gov