numerical-methods

What problem does it solve?

This Skill provides guidelines and practical techniques to identify, prevent, and diagnose numerical stability issues in statistical computations, improving accuracy and reliability of results.

Core Features & Use Cases

• Floating-point fundamentals: Understand precision, range, and machine epsilon to design robust computations.
• Stability patterns: Apply log-sum-exp, softmax stabilization, and proper variance/gradient calculations.
• Robust linear algebra: Use QR/Cholesky decompositions and pseudoinverse to avoid ill-conditioning in regression and optimization.
• Applications: Useful in estimators, optimization routines, Monte Carlo simulations, and numerical integration.

Quick Start

Run a quick check on a sample dataset to validate stability of a simple log-sum-exp, then report any instability and suggested fixes.