VegasMapping¶

class madspace.VegasMapping(self: madspace._madspace_py.VegasMapping, dimension: SupportsInt, bin_count: SupportsInt, prefix: str = '')¶

Bases: Mapping

VEGAS adaptive-sampling mapping.

A separable piecewise-linear reparametrization of the unit hypercube, one axis-aligned grid of bin_count bins per dimension, adapted to the integrand during a warm-up run [1, 2]. The grid is a compute-graph global named after prefix; call initialize_globals(context) once before use.

batch is the leading batch dimension.

Inputs

latent – float, shape (batch, dimension) – uniform coordinates in [0, 1).

Conditions

None.

Outputs

data – float, shape (batch, dimension) – the reparametrized coordinates.

In addition every mapping returns a weight (float, shape (batch,)), the Jacobian of the transformation.

References

[1] T. Heimel, O. Mattelaer, R. Winterhalder, “MadSpace”, https://arxiv.org/abs/2602.06895 (Sec. 3.2.2)
[2] G. P. Lepage, “Adaptive multidimensional integration: VEGAS enhanced”, https://arxiv.org/abs/2009.05112

Parameters:

dimension – Number of dimensions.
bin_count – Number of grid bins per dimension.
prefix – Prefix for the grid global name.

bin_count(self: madspace._madspace_py.VegasMapping) → int¶: Number of grid bins per dimension.

dimension(self: madspace._madspace_py.VegasMapping) → int¶: Number of dimensions.

grid_name(self: madspace._madspace_py.VegasMapping) → str¶: Global name of the VEGAS grid.

initialize_globals(self: madspace._madspace_py.VegasMapping, context: madspace._madspace_py.Context) → None¶: Register the grid global on context.