Smooth approximation is a mathematical technique used to approximate non-smooth functions with smooth ones, often to facilitate analysis or computation. This approach is crucial in optimization, numerical analysis, and machine learning, where differentiability is required for gradient-based methods.