Geek Answers Handbook

Recommendations for quasi-random sequences with low frustration (e.g. Sobol) in Python / SciPy?

I would like to use a quasi-random sequence, in particular Sobol, in a SciPy based simulation. Any recommendations on existing, effective packages?

+9
python scipy random
source share
4 answers

Is this project any use? If not, the corresponding Wikipedia article mentions C and Fortran routines in NAG and Numericical Recipes, which probably shouldn't be hard to wrap or override in Python. There are also C routines in GSL .

+6
source share

For Sable strings, try sobol_seq .

Generally speaking, the best package I have found for working with quasi-random sequences is diversification .

There are also packages focused on specific implementations, for example, sudoku_lhs works with Latin hypercubes and an option for restricting Sudoku type.

PyDOE implements at least a Latin hypercube (maybe more).

The most interesting package I found is py-design , which creates a wrapper for Fortran 90 codes on about 15 methods. Unfortunately, this does not seem to work (some assets are missing).

+2
source share

In the context of sensitivity analysis, the SALib library seems interesting. It has a Sobol sample generator and uses SciPy. Link here: https://salib.readthedocs.io/en/latest/

+1
source share

Chaospy is also a valid option. This allows you to choose several approaches to the selection with low mismatch (including "Sable", "Latin hypercube", etc.). See the documentation for more details.

+1
source share

More articles:


All Articles