§5.5, p. 84 — the action of E(3) leaves distances between points unchanged
ProvedGeomDL.EGNN.e3_preserves_disteuclidean-groupgeometric-deep-learningisometryp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1
Let act on by , with a orthogonal matrix and . Then for all ,
where is the Euclidean norm.
This is the fact the book invokes for the E(3) equivariance of the Satorras et al. layer: the layer sees the coordinates only through the squared distances , which rigid motions preserve.
Formalization Note ranges over Matrix.orthogonalGroup (Fin 3) ℝ, so reflections are included, and acts on EuclideanSpace ℝ (Fin 3) through Matrix.toEuclideanLin.
Preamble
import Mathlib import Definitions.Def_GeomDL_EGNN_Defs
Formal statement
namespace GeomDL.EGNN
/-- §5.5, p. 84: "the action of E(3) necessarily leaves distances between nodes unchanged":
for every orthogonal `R ∈ O(3)` and translation `b ∈ ℝ³`, `‖(Rx + b) − (Ry + b)‖ = ‖x − y‖`. -/
theorem e3_preserves_dist (R : Matrix.orthogonalGroup (Fin 3) ℝ) (b : E3) (x y : E3) :
‖rigid R b x - rigid R b y‖ = ‖x - y‖ := by sorry
end GeomDL.EGNN
Source
Bronstein et al., Geometric Deep Learning (arXiv:2104.13478v2), §5.5, p. 84, "the action of E(3) necessarily leaves distances between nodes unchanged"
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.