The minfx project is a Python package for numerical optimisation, being a large collection of standard minimisation algorithms. The name minfx is simply a shortening of the mathematical expression min f(x).

Local optimization algorithms

Line search methods

• Steepest descent.
• Back-and-forth coordinate descent.
• Quasi-Newton BFGS.
• Newton.
• Newton-CG.

Trust-region methods

• Cauchy point.
• Dogleg.
• CG-Steihaug.
• Exact trust region.

Conjugate gradient methods

• Fletcher-Reeves.
• Polak-Ribiere.
• Polak-Ribiere +.
• Hestenes-Stiefel.

Miscellaneous

• Grid search.
• Simplex.
• Levenberg-Marquardt.

Auxiliary algorithms

Step selection subalgorithms

• Backtracking line search.
• Nocedal and Wright interpolation based line search.
• Nocedal and Wright line search for the Wolfe conditions.
• More and Thuente line search.

Hessian modifications

• Eigenvalue modification.
• Cholesky with added multiple of the identity.
• Gill, Murray, and Wright modified Cholesky algorithm (GMW81).
• The Schnabel and Eskow 1999 algorithm (SE99).

Constraint algorithms

All minimization methods can be constrained by the Method of Multipliers (also known as the Augmented Lagrangian).

Projects using minfx

The following is a list of projects using the minfx library. If you would like your project added, please send a message to the minfx-devel mailing list.

• relax - Molecular dynamics by NMR data analysis.

Registration Date: Sun 04 Nov 2007 09:12:33 PM UTC
License: GNU General Public License V2 or later
Development Status: 6 - Mature