SmartGA

Purpose

SmartGA searches for compact mathematical expressions that fit supplied input-output samples. Candidate formulas are encoded as small instructions so mutation can replace individual operations.

Engineering

• Reproducible genetic search through explicit random seeds.
• Strict parsing, complete numeric-token validation, and finite scoring.
• Strong exception safety when mutating candidate expressions.
• Portable CMake packaging with an installed SmartGA::core target.
• Linux and Windows CI, warning-as-error builds, and standalone consumer tests.

Finding

Multi-seed probes separated engineering reliability from search quality: the implementation is reproducible and well-tested, but the current instruction representation does not reliably discover simple target formulas. That representation bottleneck is the next research problem rather than a result to hide behind packaging quality.

View the repository