Skip to content

Verification and comparison

Broader topicNumerical modeling

Choose numerical methods and execution paths, then assess the evidence behind a model result.

Follow the science

Verification and comparison

Read numerical comparisons and recorded qualification evidence with their limits.

Use several forms of evidence together: analytic limits, inventory checks, refinement, cross-implementation comparison, and external observations. Each rules out different errors. Performance measurements also need workload, device, synchronization, and accuracy context to be scientifically useful.

A question to investigate

What plausible error would remain undetected by your current checks, and which independent diagnostic would expose it?

Start here

Validation records

Read, calculate, investigate