Did You Know What You Just Counted? Automated Verification of Numerical Computations
summary
Describes problems with floating-point computations where every operation introduces rounding errors that accumulate and can invalidate results. Presents automatic methods to check numerical accuracy and detect when accumulated errors make computed results unreliable.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:29Floating-point numbers
- 4:29Error propagation
- 9:56Distrust in computations
- 10:17Cryptonite solution
- 13:34Matrix determinant example
- 16:08Library precision storage
- 17:40Overhead and performance
- 22:51Library development plans
- 23:57Q&A
related talks
44mWhy 0.1+0.2 != 0.3, or the Mysterious World of Floating-Point Numbers
21mSafe Numerics Library
50mRecent Advances in Floating-point (Static) Analyses
19mRevealing Floating-Point Accumulation Orders in Software/Hardware Implementations
42mfixed_point
34mProgramming’s Greatest Mistakes
24m"It's all about money, money, money"
01h 05m“Safe Numerics”