Description
The numerical stability of algorithms provides a bound on precision and reliability, where unstable algorithms will be unable to provide precise results in the face of noise or error. Monte Carlo Arithmetic (MCA) is a stochastic arithmetic technique used in the evaluation of such errors and allows for empirical estimation of numerical stability. We instrumented the scientific Python stack with MCA to present an environment that allows for the perturbed execution and evaluation of arbitrary Python code. This "Fuzzy" environment provides a fully transparent experience for users who wish to evaluate the robustness of their tools and pipelines.