Contents of the lecture
Numerical integration: simple deterministic methods in 1D; Monte Carlo integration.
- 1D classical integration methods: trapezoidal rule, Simpson...
- Monte Carlo integration in 1D: "hit or miss" (or "acceptance-rejection"); "sample mean".
- algorithms to improve the efficiency: importance sampling
- Handling errors: variance reduction with (i) average of the averages (ii) block average
Multidimensional numerical integration: comparison between deterministic and Monte Carlo methods.
Error analysis
- error in classical methods with equispaced abscissas in one and higher dimensions
- comparison with errors in Monte Carlo method
More on 1D numerical integration: Gaussian Quadrature
- Classical 1D integration: Gaussian Quadrature, Ortjogonal polynomials
- Gauss-Legendre quadrature (use of "gauleg" subroutine from Numerical Recipes)
Generalities about the use of Numerical Recipes
References:
- Chapter 1.2 "Numerical quadrature" from "Computational Physics" by Koonin; chapter 8.1-2
- Chapter 11 "Numerical Integration" from "Computer simulation Methods" by Gould-Tobochnik (II ed)