evox.metrics.igd

evox.metrics.igd#

Module Contents#

Functions#

igd

Calculate the Inverted Generational Distance (IGD) metric between a set of solutions and the Pareto front.

API#

evox.metrics.igd.igd(objs: torch.Tensor, pf: torch.Tensor, p: int = 1)[source]#

Calculate the Inverted Generational Distance (IGD) metric between a set of solutions and the Pareto front.

Parameters:
  • objs – A tensor of shape (n, m), where n is the number of solutions and m is the number of objectives. Represents the set of solutions to be evaluated.

  • pf – A tensor of shape (k, m), where k is the number of points on the Pareto front and m is the number of objectives. Represents the true Pareto front.

  • p – The power parameter used in the calculation (default is 1). This defines the distance metric (L^p norm).

Returns:

The IGD score, a scalar representing the average distance of the solutions to the Pareto front.

Note:

The IGD score is lower when the approximation is closer to the Pareto front.