paref.pareto_reflections.find_edge_points#
Classes
|
Find the edge points of the Pareto front |
- class paref.pareto_reflections.find_edge_points.FindEdgePoints(dimension_domain: int, dimension: int, epsilon: float = 0.001)[source]#
Bases:
MinimizeWeightedNormToUtopiaFind the edge points of the Pareto front
Warning
This Pareto reflection assumes that there exist edge points
When to use#
This Pareto reflection should be used if the edge points of the Pareto front are searched.
Note
In to dimensions, the edge points of the Pareto front always exist.
What it does#
The Pareto points of this map are the ones which minimize the weighted sum where one component is given much smaller weight than the others.
Mathematical formula#
\[p(x) = \sum_{i=1,...,n,i\neq j}\epsilon x_{i}+ x_j\]where \(j\) is the component in which the minimum is searched.
Examples
# TBA: add
Specify the dimension of the input domain and the component of which the edge point is searched
Warning
The smaller epsilon, the better. However, picking an epsilon too small may lead to an unstable optimization.
- param dimension_domain:
dimension of domain (i.e. dimension of target space of blackbox function)
- type dimension_domain:
int
- param dimension:
component of which the edge point is searched
- type dimension:
int
- param epsilon:
weight on the component
- type epsilon:
float default 1e-3
- property dimension_codomain: int#
- property dimension_domain: int#