SearcharxivSearch

arXiv subjects

Irmi Schneider

Publications and source records attributed to Irmi Schneider.

2 recordsLinked to original sources

Metric Rarity and the Emergence of Symmetry in G-Invariant Potential Surfaces

Let X be an irreducible complex affine algebraic variety defined over $\mathbb{R}$, equipped with a faithful action of a finite group G, and let Y = X // G denote the categorical quotient with projection $π$. We study the geometry of the real image $L = π(X(\mathbb{R})) \subset Y(\mathbb{R})$ and its consequences for G-invariant optimization. Equipping $Y(\mathbb{R})$ with the measure induced by a G-invariant metric on X, we prove that the relative volume of L in $Y(\mathbb{R})$ equals $(\#\mathrm{Inv}(G))^{-1}$, where $\mathrm{Inv}(G)$ is the set of involutions of G. For the symmetric group $S_n$ acting on $\mathbb{R}^n$, this ratio decays super-exponentially in n. In particular, L is metrically rare within the ambient real quotient. We apply this result to two phenomena observed in G-invariant optimization problems: Regime I (Rarity of asymmetric critical points). The super-exponential decay of the volume of L renders the interior $L^\circ$ statistically negligible as a locus for critical points. This geometric rarity provides a rationale for the observed prevalence of symmetry: generic critical points are constrained to the boundary strata of L, corresponding to orbits with non-trivial stabilizers. Regime II (Energetic ordering by symmetry). We formulate the Active Constraint hypothesis: due to the metric rarity of the real image L, the landscape is dominated by a global gradient that drives the deepest descent trajectories toward the boundary of L. This global gradient directs the global minimum into the high-codimension strata of the boundary -- corresponding to large stabilizers -- thereby establishing a structural link between low energy and non-trivial stabilizers. This mechanism rationalizes the funnel topography of Lennard-Jones clusters, where the system is funneled into a crystallized ground state.

math.OC

Ubiquitous Symmetry at Critical Points Across Diverse Optimization Landscapes

Symmetry plays a crucial role in understanding the properties of mathematical structures and optimization problems. Recent work has explored this phenomenon in the context of neural networks, where the loss function is invariant under column and row permutations of the network weights. It has been observed that local minima exhibit significant symmetry with respect to the network weights (invariance to row and column permutations). And moreover no critical point was found that lacked symmetry. We extend this line of inquiry by investigating symmetry phenomena in real-valued loss functions defined on a broader class of spaces. We will introduce four more cases: the projective case over a finite field, the octahedral graph case, the perfect matching case, and the particle attraction case. We show that as in the neural network case, all the critical points observed have non-trivial symmetry. Finally we introduce a new measure of symmetry in the system and show that it reveals additional symmetry structures not captured by the previous measure.

cs.LG