The affine-loss case has zero infimum and no minimizing pair
ProvedPGDRevision20261113.affine_loss_nonattainmentbregman-geometryconvex-optimizationformalization
Take and . In the interval family let with ; in the real-line family it suffices that and . For , define the usual error using on the interval, or on the line.
Every feasible pair has strictly positive error norm. Quadratic geometry satisfies
and for every some makes this norm smaller than . Consequently the global infimum is zero and no feasible pair attains it.
Related results. Exact profile performance (Theorem 4.1) · Geometry classification (Theorem 5.1)
Preamble
import Definitions.Def_PGDRevision20261113_Examples open PGDRevision20261113
Formal statement
theorem PGDRevision20261113.affine_loss_nonattainment : AffineLossNonattainmentClaim := by sorry
Source
Author manuscript, Exact Residual Bounds and Optimal Geometry Selection for Proximal Gradient Methods, revised reading edition, 13 November 2026; Section 5, paragraph following the classification proof; sections/05_geometry_classification.tex, For the excluded affine case.