This library implements an algorithm for learning the boundary between an upward-closed set X1 and its downward-closed component X2 (i.e., X=X1+X2). Generally, the library supports spaces X of dimension N. The algorithm selects sampling points x=(x1,x2,...,xN) for which it submits membership queries ’is x in X1 ?’ to an external Oracle. Based on the Oracle answers and relying on monotonicity, the algorithm constructs an approximation of the boundary, called the Pareto front. The algorithm generalizes binary search on the continuum from one-dimensional (and linearly-ordered) domains to multi-dimensional (and partially-ordered) ones. Applications include the approximation of Pareto fronts in multi-criteria optimization and parameter synthesis for predicates where the influence of parameters is monotone. This library is based on the work-in-progress paper [1]. [1] [Learning Monotone Partitions of Partially-Ordered Domains (Work in Progress) 2017.〈hal-01556243〉] (https://hal.archives-ouvertes.fr/ha...)
ParetoLib
- Software distribution and licensing: Open-source distribution
- Software maturity: Basic usage works, terse documentation.
- Software Maintenance and Evolution: Active maintenance, plans for future evolution
Multidimensional Pareto Boundary Learning Library for Python
Voir en ligne : ParetoLib Tool
