Papers

We establish the first universal upper bounds for the real degree of the contour of the amoeba of a smooth complete intersection in the algebraic torus. Our approach extends the Pfaffian method of Lang–Shapiro–Shustin from hypersurfaces to arbitrary codimension by introducing a logarithmic conormal framework based on the logarithmic conormal bundle, the logarithmic Grassmann map, and determinantal rank conditions. We prove that, on suitable logarithmic conormal charts, the critical locus is locally defined by Schur–complement equations arising from the logarithmic conormal matrix. This yields explicit universal contour-degree estimates in both regimes $n\ge 2r$ and $n<2r$. We further replace total-degree arguments by Bernstein's theorem to obtain sparse mixed-volume bounds determined by the Newton polytopes of the transformed equations. These estimates are frequently much sharper than the corresponding universal bounds and provide a higher-codimensional analogue of the Lang–Shapiro–Shustin theory together with a new geometric interpretation of amoeba contours through logarithmic conormal geometry.

This paper presents a proof-of-concept workflow for equation-oriented process optimization that runs entirely on a GPU. Process optimization models often incorporate complex interconnected unit operations, dynamics, and uncertainties, resulting in large nonlinear programs that can be computationally demanding for conventional CPU-based solvers. Although emerging GPU-based solvers offer substantial computational benefits, their application to process optimization has been limited by the lack of GPU-compatible process modeling tools. We address this gap by prototyping the GPU-compatible process optimization models using an existing GPU-capable optimization software stack, including ExaModels (algebraic modeling system), MadNLP (optimization solver), and cuDSS (linear solver). ExaModels formulates the process optimization problem in a GPU-compatible way by exposing its repeated algebraic structure, while MadNLP and cuDSS solve the resulting nonlinear program on the GPU. This workflow is demonstrated on a CO2 absorber design problem under feed uncertainty, in which a shared column diameter is minimized subject to equilibrium and hydraulic constraints in all scenarios. For the largest case with 5,000 scenarios and 1.5 million variables, the GPU workflow achieves a speedup of approximately 21\times over a single-threaded CPU baseline using JuMP, Ipopt, and MA57.

We show that there is a constant $c>0$ such that, for all sufficiently large $N$, there is a subset $A \subseteq \{1,\dots,N\}$ of size $>cN$ such that for any two distinct elements $a,b$ in $A$, the average of $\frac{1}{a}$ and $\frac{1}{b}$ is not a unit fraction, negatively answering a question of Erdős and Graham. This also gives the best known lower bounds on the maximum size of a set of unit fractions without non-trivial three-term arithmetic progressions.

Multi-objective learning (MOL) aims to optimize multiple objectives simultaneously. The multi-gradient descent algorithm (MGDA) constructs a common descent or conflict-avoidant (CA) direction by solving a gradient-aggregation subproblem. In stochastic settings, the vanilla stochastic MGDA method, SMG, replaces full-batch gradients with mini-batch estimates, causing bias in the resulting CA direction. Existing analyses control this error using continuity of the CA-direction mapping, but do not distinguish how its sensitivity varies with the geometry of the MGDA subproblem. Motivated by this limitation, we show that the MGDA direction map is Lipschitz continuous under reduced nondegeneracy, whereas its general \(1/2\)-Hölder exponent on bounded sets is sharp. This geometry motivates the Multi-objective Regularity-aware (MoRe) method, which retains adaptive MGDA aggregation in the regular regime and switches to a fixed-weight linear scalarization (LS) near degeneracy. The fallback branch skips the MGDA subproblem solve, which can reduce aggregation cost on fallback iterations while avoiding a potentially noise-sensitive update. Theoretically, we establish a per-iterate CA guarantee, and a \(\widetilde{\mathcal O}(T^{-1})\) convergence rate, improving upon the previously established \(\widetilde{\mathcal O}(T^{-1/4})\) SMG guarantee, where \(\widetilde{\mathcal O}\) hides logarithmic factors. Experiments demonstrate favorable multi-task performance and convergence behavior consistent with the theoretical rate.

In this paper, we study a system of nonlocal conservation laws motivated by traffic flow: a nonlocal version of the generalized Aw-Rascle-Zhang (GARZ) model. The nonlocality arises from downstream spatial averaging of the velocity by a one-sided kernel. We prove the existence and uniqueness of weak solutions for initial data of bounded variation via a fixed-point argument in the nonlocal velocity. We also establish stability with respect to the initial datum and an approximation of weak solutions by strong solutions in $L^1$. Under additional, physically meaningful assumptions on the velocity and the initial datum, we obtain either a maximum principle for the density or invariant-region estimates. Finally, we study the singular limit as the nonlocal kernel converges to a Dirac distribution. Indeed, under additional assumptions, convergence to the unique local entropy solution can be proved. Some numerical simulations are provided, and the paper concludes with a discussion of open problems.

Let $\mathcal O$ be a special nilpotent orbit in the Lie algebra $\mathfrak g$ of a simple algebraic group $G$. We give two proofs of the result that every special piece ${\mathcal P}(\mathcal O)$ in $\mathfrak g$ is the quotient of a smooth $G$-variety $X$ by the action of a certain finite group $H$. We first deduce the result from a similar result for transverse slices, established in earlier work of the first three authors and Fu. Then we give a more explicit construction of $X$, as a subvariety of the closure of a $G$-orbit in the direct sum of $\mathfrak g$ and some fundamental weight representations of $G$. Both methods apply to classical $\mathfrak g$, where we give new proofs of this result, which was first proved by Kraft and Procesi. The result in the exceptional groups was conjectured by Lusztig. Our first proof shows that there can be several $G$-varieties $X$ that satisfy the conjecture, related to a natural embedding of $H$ in the fundamental group of $\mathcal O$. In an appendix, we relate this natural embedding to Lusztig's definition of $H$ that arises from the family in the Weyl group of $G$ attached to $\mathcal O$ and from the Springer correspondence.

Interleaving mitigates burst errors but introduces decoding delay and removes temporal error structure that a channel-aware decoder could exploit. We consider packet-level selection between a random linear code and the same code used with cross-codeword interleaving, over a channel with an unknown number of on/off interferers. The receiver uses Guessing Random Additive Noise Decoding (GRAND) with a replaceable noise model and feeds aggregate channel statistics back to a Bayesian estimator at the transmitter. Once the interference amplitudes and timing parameters are estimated, the receiver's noise model is replaced: it computes hidden-Markov-model posterior bit-flip probabilities and uses them to order GRAND queries. A discounted Thompson sampler selects between the two transmission modes using a goodput-minus-latency reward whose distribution is endogenously nonstationary: receiver adaptation, rather than channel change, alters the value of each mode. Across five simulation seeds, the interleaved mode is preferred before channel estimation converges. After the learned decoder is activated, the non-interleaved mode becomes preferable because it achieves lower block error rate without interleaving delay. In the reference configuration, the learned noise model reduces block error rate by approximately one order of magnitude relative to ORBGRAND. Using partial channel estimates before full convergence reduces pre-convergence block error rate by up to $4.5\times$. Adding model-predicted utilities as confidence-weighted pseudo-observations reduces post-transition selection of the inferior arm by approximately $65\%$. Under an idealized airtime conversion at a 100~MHz 5G~NR-like symbol rate, the learning transient corresponds to a few milliseconds of occupied symbol time.

The following notes are my attempt to clarify, to myself, the OpenAI proof of the Cycle Double Cover Conjecture. This write-up offers no new insights, but some people may find it easier to read.

We study the minimum arclength of spherical $t$-design curves, i.e., closed rectifiable curves on $S^d$ whose normalized arclength measure exactly integrates every polynomial of degree at most $t$. We prove an explicit spectral lower bound that is sharp for $t=1$ in all spheres and for $t=2$ in every odd-dimensional sphere, yielding the first exact optimality results for spherical $t$-design curves with $t>1$. For even-dimensional spheres, we construct $2$-design curves whose lengths asymptotically match the lower bound as $d\to\infty$, and in $S^2$, we use numerical optimization and the calculus of variations to derive a candidate for the shortest $2$-design curve.

Brändén and Huh showed that Lorentzian polynomials unify Hodge-Riemann relations in combinatorics: their supports are M-convex, and every M-convex set supports a Lorentzian polynomial. Baker, Huh, Kummer, and Lorscheid later proved that, for every $q>0$, the projectivized space $\mathbf{P}\operatorname{L}_J$ of Lorentzian polynomials with support $J$ is homeomorphic to the thin Schubert cell $\operatorname{Gr}^{\mathrm{w}}_J(\mathbb{T}_q)$ of weak representations of $J$ over the generalized triangular hyperfield $\mathbb{T}_q$. We study the quantitative relation between Lorentzian polynomials and representations over triangular hyperfields. For every matroid $M$, we prove that some $q>0$ depending on $M$ satisfies $\operatorname{Gr}^{\mathrm{w}}_M(\mathbb{T}_q)\subseteq\mathbf{P}\operatorname{L}_M\subseteq\operatorname{Gr}^{\mathrm{w}}_M(\mathbb{T}_2)$. Thus $\mathbf{P}\operatorname{L}_M$ lies between two thin Schubert cells, each homeomorphic to it. More generally, for every M-convex set $J$, some $q>0$ depending on $J$ satisfies $\operatorname{N}\operatorname{Gr}^{\mathrm{w}}_J(\mathbb{T}_q)\subseteq\mathbf{P}\operatorname{L}_J\subseteq\operatorname{N}\operatorname{Gr}^{\mathrm{w}}_J(\mathbb{T}_2)$, where $\operatorname{N}$ denotes normalization. We also study $q(M):=\sup\{q>0:\operatorname{Gr}^{\mathrm{w}}_M(\mathbb{T}_q)\subseteq\mathbf{P}\operatorname{L}_M\}$. For $q(n):=q(U_{2,n})$, we prove $q(4)=2$ and $q(5)=\log_2 3$, with matching upper and lower bounds of order $1/n$; hence $q(n)=\Theta(1/n)$, so in particular no universal positive lower bound for $q(n)$ exists.

Let $K = k(X)$ be the function field of a smooth geometrically integral variety $X$ of dimension $\geq 2$ over a field $k$ of characteristic 0 and $V$ be the set of discrete valuations of $K$ associated with the prime divisors on $X$. We show that if $D$ is a $k$-defined group of multiplicative type, then the corresponding Tate-Shafarevich group $Sha(D,V) = \ker \left(H^1(K,D) \to \prod_{v \in V} H^1(K_v, D) \right)$ is finite in the following situations: (1) $k$ is finitely generated and $X(k) \neq \emptyset$; (2) $k$ is a number field. This complements previous work of Harari and Szamuely, which considered the case where $X$ is a curve.

In this paper, we define several new two-variable elliptic genera for even dimensional spin manifolds which are the index for some Dirac operators and holomorphic $SL(2,Z)$, $\Gamma_0(2)$, $\Gamma^0(2)$, $\Gamma _{\theta}$-Jacobi forms. By these Jacobi forms, we can get some $SL(2,{\bf Z})$ and $\Gamma^0(2)$ modular forms. By these $SL(2,{\bf Z})$ and $\Gamma^0(2)$ modular forms, we get some interesting anomaly cancellation formulas for spin manifolds.

We prove that the longer and shorter Sackin indices of a uniformly random simplex tree-child network with $n$ taxa admit joint distributional limits after rescaling by $n^{-7/4}$. The limiting distributions are described by functionals of a Brownian excursion. We also identify the limiting law of the height after rescaling by $n^{-3/4}$, thereby answering a question of Zhang~(2022). Moreover, we establish sharp tail bounds for the height, which imply convergence of all moments in the above distributional limits. We further obtain a scaling limit for the entire height profile of the leaves. Finally, we determine the local limits of large simplex networks around the fixed root, a uniformly random vertex, and a uniformly random leaf.

We investigate the computational complexity of bilevel integer linear programming. While Jeroslow~(1985) established that the decision version of this problem is $\Sigma^p_2$-complete when restricted to binary variables, we prove that this $\Sigma^p_2$-completeness persists even for general integer variables, settling a question that remained open for over 40 years. Furthermore, we analyze the impact of various structural assumptions on computational complexity. Notably, we strengthen the result of Köppe et al.~(2010) by proving polynomial-time solvability whenever the total number of upper- and lower-level variables is fixed, without any additional assumptions.

We study the complete intersection property of projective closures of affine monomial curves. For monomial curves in the four-dimensional affine space, we give a complete characterization of when the defining ideals of their projective closures are complete intersections. We also investigate when the projective closures are arithmetically Cohen–Macaulay.

We develop a unified method for transforming and evaluating infinite series involving products of terms of second-order linear recurrences in the denominator. The approach is based on a discrete Abel-type summation formula (summation by parts), which converts reciprocal series with three or more factors into expressions exhibiting a partial telescoping structure. As a consequence, we obtain general transformation formulas for series of the form $\sum\limits_{k=1}^{\infty} \frac{(\pm 1)^k}{w_{rk+l} w_{mk+s} w_{m(k+1)+s}}$, together with extensions to products of four or more terms. These formulas provide a systematic framework that unifies and extends many known identities for Fibonacci and Lucas numbers. In addition, the method leads to explicit evaluations and identities involving several classical combinatorial sequences, including Catalan numbers, harmonic numbers, and Stirling numbers of both kinds. A key feature of the approach is that it naturally distinguishes between even and odd values of the parameter $m$, leading to structurally different representations. The results show that summation by parts is an effective and flexible tool for reducing multi-factor reciprocal sums to simpler forms.

Li and Boyle showed that the Penrose tiling defines a quantum error-correcting code: superpositions of tilings over isometry orbits protect quantum information against erasure of any bounded region. We extend the construction to the aperiodic monotiles discovered by Smith, Myers, Kaplan and Goodman-Strauss. For the Hat, we prove strong local indistinguishability for all Hat tilings, and we prove local recoverability unconditionally for all nonsingular Hat tilings via the torus parametrization of the underlying cut-and-project scheme. The remaining singular case reduces to one sharply posed geometric question – can a region that is a union of hats be retiled a second way? – which we verify computationally has no counterexample up to a substantial scale: a certified $2490$-tile patch admits precisely one tiling by hats, so all $2^{2490}$ of its tile-subregions retile uniquely. Unlike the Penrose, Ammann-Beenker and Fibonacci tilings, both monotiles form two local-indistinguishability classes, so their code spaces split into two erasure-correcting sectors carrying a superselected classical label. Whether the label survives depends on which isometries are gauged: the Spectre's classes are exchanged by a $30^{\circ}$ rotation and merge once all proper isometries are gauged, whereas the Hat's are exchanged only by reflections. Under the physically natural gauge group $SE(2)$, the Hat code therefore stores one robust classical bit – the handedness of its long-range order, readable in any bounded window with separation $\Delta = |K|\sqrt{5}/3$ – alongside its protected quantum sectors. It is the reflexible monotile, not the chiral one, that carries the chirality bit. We give the exact Perron-Frobenius data of both codes, including the per-class reflected-Hat frequencies $(3\mp\sqrt{5})/6$ and the Spectre orientation-class frequencies $(5\pm\sqrt{15})/10$.

Fuzzy matrices provide an effective framework for modeling uncertainty in scientific and engineering problems, particularly fuzzy linear systems. This work transforms a general FLS into a crisp linear system using an embedding approach and reduces it to a standard block structured form via column operations. A direct LU decomposition is developed under suitable range conditions, enabling the computation of minimum norm solutions of rectangular FLS using the generalized inverses. A full rank decomposition is further proposed to compute the Moore Penrose inverse for arbitrary rectangular matrices, and consistency conditions for these inverses are established. A unified framework for obtaining strong fuzzy solutions based on monotonicity and non-negativity constraints are presented. Efficient algorithms based on LU, QR, and SVD decompositions are developed for the block structured matrix. The applicability and computational efficiency of the proposed methods are demonstrated through fuzzy circuit equations and Markov chain processes.

The snake-in-the-box problem asks for a longest induced path in the hypercube graph $Q_n$. We find a length-191 snake in dimension $n=9$, the lowest dimension where the maximum is unknown, improving the previous record of 190 that had stood for 14 years. We also establish new lower bounds in dimensions 10-13. To find these records, we introduce snakepits, collections of disjoint snakes, to expand the search space and open new routes between snakes. This motivates our new Snakepit-in-the-Box benchmark, which seeks maximal edge counts when allowing multiple components. Finally, we introduce Beam Anchor, a search-supervised learned constructor algorithm that finds 100 inequivalent length-190 snakes in dimension 9.

We prove, for all fixed $0 < \delta < 1$, and all sufficiently large $n$, that there exists $S \subset [n]$ with $|S| \ge \delta n$ such that $A + B \not \subset S$ for all ${A, B \subset \mathbb{N}}$ satisfying $$\min\big\{|A|, |B|\big\} \ge \big(3 + o(1)\big) \frac{\log n }{ \log (1 / \delta)}.$$ A very recent result of Hernández and Hetzel shows that our bound is sharp up to a factor of 3, and together our results settle a conjecture of Kra, Moreira, Richter, and Robertson. In fact, we prove that a $\delta$-dense random subset of $[n]$ is a valid choice for $S$ with high probability, and that one can take $n^{-\alpha} \le \delta \le 1 - c$ where $c > 0$ is fixed and $\alpha > 0$ depends only on the $o(1)$ error, answering another question of the same authors in a strong form.

A set of points $C \subset \mathbb{R}^n$ is called canonically Ramsey if there is some set of points $S\subset \mathbb{R}^{n'}$ such that any colouring of $S$, using any number of colours, must contain either a monochromatic copy of $C$ or a rainbow copy of $C$. Mao, Ozeki, and Wang introduced this notion, showing that 30-60-90 triangles are canonically Ramsey. Since then, various other canonically Ramsey configurations have been identified. The author showed that cuboids are canonically Ramsey, while Ge, Shu, Xu, and Yu recently showed that simplices are canonically Ramsey. We extend both of these results, proving that all products of simplices are canonically Ramsey.

In the last decades, spacetime noncommutativity and the associated deformations of relativistic symmetries have attracted a lot of interest, as several phenomenological windows into quantum gravity are approaching genuine Planck-scale sensitivity. However, the physical significance of the mathematical structures introduced to deal with spacetime noncommutativity is still debated, and some crucial pieces are missing or poorly understood. From time to time it has been suggested that valuable insight into these conceptual challenges could come from the analysis of appropriately designed first-quantized toy models, in which major technical and interpretive issues can be effectively managed or sidestepped altogether. In this paper we take seriously this suggestion and develop the first fully consistent and fully relativistic first-quantized model of two particles propagating and interacting on a noncommutative spacetime. The specific spacetime noncommutativity implemented in our model, which we call 'time-commutative $\kappa$-plane', had already been proposed as a suitable arena for a first-quantized analysis, but previous studies were mostly heuristic and failed to provide a full description of the deformed relativistic symmetries. We here go much beyond those pioneering attempts: we find a full characterization of the appropriate deformed Poincaré symmetry algebra as well as its Galilean limit; we build a single-particle quantum model carrying an irreducible representation of the deformed Galilei algebra; we exhibit two consistent, Galilean-relativistic descriptions of a system of two quantum particles interacting via a deformed harmonic potential, finding that the structure of the two-particle symmetry generators is intimately connected with the deformation of the interaction law.

Computer-assisted proofs of self-similar singularity formation for fluid equations often rely on numerically constructed approximate profiles. One effective approach to establishing stability of perturbations around a numerically constructed profile is to perform weighted energy estimates with singular weights near the singularity. However, the weighted norms require exact local vanishing conditions that are not automatically preserved by the equations nor the numerical construction. In this paper, we review an analytic low-rank correction method first developed in [ChenHou2023a,ChenHou2023b] to overcome this difficulty. The numerical step determines coefficients, rigorous bounds, and low-order defect modes in explicit global basis representations, while the required vanishing conditions are enforced analytically through low-rank corrections derived from Taylor expansions of the relevant quantities represented in a smooth basis. For completeness, we briefly review the singularly weighted estimates and a quantitative finite-rank perturbation method in the 2D Boussinesq / 3D Euler stability argument, where singular weights and the required vanishing order arise. Against this background, we formulate the local correction principle in a simplified setting, explain the correction of the residual error in numerical constructions of approximate space-time solutions and the stream function, and discuss its broader applicability to computer-assisted stability analysis for nonlocal PDEs.

We show, using group theoretic constructions, that the system of elliptic partial differential equations whose solutions determine the constant mean curvature immersions into hyperbolic 3 space are equivalent to the elliptic Liouville equation. In particular, the Darboux integrability of each equation shows that each of these equations admit an equivalent quotient representation which descends to an equivalence of the equations. The explicit map providing the equivalence is given.

The “goldfish” equations, so named because of their striking beauty, are a system of $N$ nonlinear ODE's, where $N$ is an arbitrary integer. They can be solved exactly in a very simple manner, by transforming them to free motion using a transformation involving the transition from the set of coefficients\/ of a polynomial to that of its zeroes. This paper aims to explore the possibility of extending this solution to the case in which $N$ is infinite. The main difficulty involves the transition from polynomials to entire functions. Another approach using non-standard analysis, is left to future work.

The undamped Duffing oscillator is a nonlinear dynamical system with broad applications in physics, engineering and biological system. We present a comprehensive analysis of this system using the Lindstedt Poincare method (LPM) and its modifications and make comparison with numerical solution obtained using higher order Runge-Kutta. It is also shown the method suggested in this article converges better than the standard LPM and Lindstedt Poincare method with Burton's modification.

We develop data-driven algorithms for maintaining $N$ independent identical machines under a block replacement policy, in which each machine is replaced upon failure and all machines are jointly replaced at regular intervals of length $k$. The goal is to learn the cost-minimizing interval $k^*$ from operational data when the lifetime distribution is unknown. At each decision epoch, the operator selects $k \in \{1, 2, \ldots, K\}$, observes the resulting failure history (a mixture of complete and right-censored lifetimes) and incurs a per-unit-time cost governed by the renewal function. We formulate this as a stochastic multi-armed bandit and propose Hoeffding- and Bernstein-based lower-confidence-bound algorithms achieving $O(K \log T)$ regret, matching the Lai–Robbins lower bound. Exploiting a nested observation property unique to block replacement, correlated variants attain $O((K-k^*)\log T)$ regret and require only $O(1)$ direct pulls of suboptimal arms $k < k^*$. A complementary Kaplan–Meier renewal algorithm estimates the lifetime distribution nonparametrically from censored data, achieving almost-sure policy consistency and empirically near-zero incremental regret at long horizons. We additionally analyze two average-cost MDPs: a time-elapsed formulation establishing that block replacement is optimal within its policy class for any lifetime distribution, and an age-vector formulation proving a monotone threshold structure under increasing failure rate distributions and providing a gold-standard cost benchmark. Numerical experiments confirm the theoretical ordering and reveal structural cost gaps between optimal block and age-dependent replacement.

In this paper, we establish an analogue of the Kővari-Sós-Turán Theorem for $\operatorname{GF}(q)$-representable matroids. For $2\leq s\leq t$, we show that if $M$ is a rank-$n$ simple $\operatorname{GF}(q)$-representable matroid having no $M(K_{s,t})$-restriction, then \[ |E(M)|=O_{q,s,t}\bigl(q^{(1-1/s)n}\bigr). \] In particular, we prove that the maximum number of elements in a simple rank-$n$ binary matroid with no $M(K_{2,t})$-restriction is $\Theta_{t}(2^{n/2})$ where the lower bound is obtained using binary Sidon sets.

We characterize bounded operators that are compact (respectively, Schatten-class) perturbations of scalar multiples of partial isometries with finite-dimensional kernel. Our characterizations are formulated in terms of the essential spectrum of $T^*T$, absolutely norm attaining operators, and the Moore-Penrose inverse. In particular, we show that an operator $T$ is a Schatten-class perturbation of a partial isometry with finite-dimensional kernel if and only if $\sigma_{\mathrm{ess}}(T^*T)$ is a singleton and the discrete spectrum of $T^*T$ satisfies a corresponding $\ell^p$-summability condition. We further obtain equivalent criteria involving the compactness (or Schatten-class membership) of $\alpha I-T^*T$ and $\alpha T^\dagger-T^*$. As applications, we establish characterizations of compact and Schatten-class perturbations of isometries, describe the corresponding behavior of Moore–Penrose inverses, and derive factorization results for closed-range operators. In particular, we provide a new Moore–Penrose inverse proof of a theorem of Şerban and Turcu and obtain an explicit formula for the factorizing operator.

We study asymptotic versions of the Brill-Noether existence conjecture on graphs via techniques inspired by the geometry of numbers. We confirm an asymptotic version of the conjecture at (and near) the half-canonical degree in several well-connected families of graphs. They include expander graphs of even valence, almost-Ramanujan graphs of a fixed valence at least five and certain random graphs. In particular, for any fixed $k \geq 5$, almost all simple, connected, $k$-regular graphs satisfy the Brill-Noether existence conjecture at the half-canonical degree up to a constant factor. The key tool is a Cheeger-style inequality for the covering radius of a certain periodic set with respect to the energy quadratic form associated with the graph. As an application, we lower bound the diameter of graphs associated with certain dynamical systems called reversal systems. We conclude with a suggestion to tackle the asymptotic version of the conjecture, in general, i.e. beyond half-canonical degrees.