MomentumPreprocessing

class madspace.MomentumPreprocessing(self: madspace._madspace_py.MomentumPreprocessing, particle_count: SupportsInt)

Bases: FunctionGenerator

Turns event momenta into invariant features for ChannelWeightNetwork.

Computes \((p_\mathrm{T}, \eta, \phi)\) for each particle together with the momentum fractions, a representation the channel-weight network can consume directly [1].

batch is the leading batch dimension.

Arguments

• momenta – float, shape (batch, particle_count, 4) – the event momenta.

• x1, x2 – float, shape (batch,) – the parton momentum fractions.

Returns

• result – float, shape (batch, output_dim()) – the feature vector.

References

• [1] T. Heimel, O. Mattelaer, R. Winterhalder, “MadSpace”, https://arxiv.org/abs/2602.06895

Parameters:

particle_count – Number of external particles.

output_dim(self: madspace._madspace_py.MomentumPreprocessing) → int

Length of the feature vector.