Talk Better numerics with SciPy
Performance in numerical optimization and linear algebra
This tutorial explores two comon classes of numerical computing problems that arise in many applications:
- Mathematical optimization, i.e. finding numerically the minimum of an objective function or cost,
- linear algebra, that encompases a variety of matrix and vector operations
The tutorial will be a blend of elementary notions in applied maths, useful for solving proactical problems, and their implementation using SciPy, with an eye on performance. It will only require an elementary level in math.
Bellow is a tentative outline that might be modified later.
Note here we are all discussing optimization on real numbers, and not e.g. integers.
- Convex versus non-convex optimization
- Smooth and non-smooth problems
- Noisy versus exact cost functions