Peer Revenue Returns subject to peer review

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Mathematics

math Watch Mathematics (sign in to use)

Wednesday 25 March

Functional Analysis

Weak limit semigroup in operator theory and ergodic theory

We study the weak limit semigroup of an operator $T$, i.e., the set of all operators being weak limit points of the powers of $T$, in three different but related contexts: Koopman operators of measure-preserving transformations, contractions/isometries/unitaries on separable Hilbert spaces and positive operators on $L^p$-spaces. Hereby we focus on finding large subsets of the weak limit semigroup, in particular in the generic case.
Read the full abstract

Probability

Channel Duality and Midpoint Invariance in Random Walks with Resetting

We study absorption of a biased nearest-neighbor random walk on a finite interval that is reset to its initial position. The central contribution is an exact duality between the two absorption channels, which holds mode by mode in the spectral representation and step by step in time, together with the invariance it produces: when the interval has even length, the ruin probability from its midpoint equals the classical reset-free value for every law of the reset times independent of the walk, provided absorption occurs almost surely. Away from the midpoint, the same duality shows that every such resetting mechanism favors the nearer boundary. For geometric resetting we obtain the ruin probability in closed form and prove that its response to the resetting probability changes sign at the midpoint and nowhere else, for every bias; when the interval has odd length the sign change falls between the two central sites. For a diffusion with drift the midpoint invariance holds for every resetting mechanism, and under Poissonian resetting the sign change is again unique. Channel duality is thus the mechanism behind the invariance and behind its robustness to the resetting protocol and to the passage to the continuum. All analytic results are checked against an independent linear solve and against simulation.
Read the full abstract

Optimization and Control

Structure, Analysis, and Synthesis of First-Order Algorithms

First-order optimization algorithms can be interpreted through the lens of dynamical systems as the interconnection of linear systems and a set of subgradient nonlinearities. This dynamical systems formulation allows for the analysis and synthesis of optimization algorithms by solving robust control problems. In this work, we use the celebrated internal model principle in control theory to structurally factorize convergent composite optimization algorithms into suitable internal models and core subcontrollers. As the key benefit, we reveal that this permits us to synthesize optimization algorithms even if information is transmitted over networks featuring dynamical phenomena such as time delays, channel memory, or crosstalk. Design of certified linearly-convergent algorithms is achieved under bisection either through a nonconvex local search or by alternation of convex semidefinite programs. We demonstrate factorization of existing optimization algorithms and the automated synthesis of new optimization algorithms in the networked setting.
Read the full abstract

All authors Jared Miller, Carsten Scherer, Fabian Jakob, Andrea Iannelli

Rings and Algebras

Structure Constant Formulas for the Universal Enveloping Algebras of the Nilpotent Lie Algebras of Dimension Five and Less

Libor Šnobl and Pavel Winternitz classified all of the Lie algebras of dimension six and smaller. Using this classification, we formulated and proved structure constant formulas for the universal enveloping algebras of the nilpotent Lie algebras of dimension five and less.

Statistics Theory · Cross-listed from stat.ME

Improving online FDR procedures via online analogs of e-closure and compound e-values

In many scientific applications, hypotheses are generated and tested continuously in a stream. We develop a framework for improving online multiple testing procedures with false discovery rate (FDR) control under arbitrary dependence. Our approach is two-fold: we construct methods via the online e-closure principle, as well as a novel formulation of online compound e-values that is defined through donations. This yields strict power improvements over state-of-the-art e-value and p-value procedures while retaining FDR control. We further derive algorithms that compute the decision at time $t$ in $O(\log t)$ time, and we demonstrate improved empirical performance on synthetic and real data.
Read the full abstract

Information Theory

Algebraic Expander Codes

Expander (Tanner) codes combine sparse graphs with local constraints, enabling linear-time decoding and asymptotically good distance–rate tradeoffs. A standard constraint-counting argument yields the global-rate lower bound $R\ge 2r-1$ for a Tanner code with local rate $r$, which gives no positive-rate guarantee in the low-rate regime $r\le 1/2$. This regime is nonetheless important in applications that require algebraic local constraints (e.g., Reed–Solomon locality and the Schur-product/multiplication property). We introduce Algebraic Expander Codes, an explicit algebraic family of Tanner-type codes whose local constraints are Reed–Solomon and whose global rate remains bounded away from $0$ for every fixed $r\in(0,1)$ (in particular, for $r\le 1/2$), while achieving constant relative distance. Our codes are defined by evaluating a structured subspace of polynomials on an orbit of a non-commutative subgroup of $\mathrm{AGL}(1,\mathbb{F})$ generated by translations and scalings. The resulting sparse coset geometry forms a strong spectral expander, proved via additive character-sum estimates, while the rate analysis uses a new notion of polynomial degree and a polytope-volume/dimension-counting argument.
Read the full abstract

Statistics Theory · Cross-listed from econ.EM

Refined Cluster Robust Inference

It has become standard for empirical studies to conduct inference robust to cluster dependence and heterogeneity. With a small number of clusters, the normal approximation for the $t$-statistics of regression coefficients may be poor. This paper tackles this problem using a critical value based on the conditional Cramér-Edgeworth expansion for the $t$-statistics. The proposed critical value guarantees third-order refinement, and it does not require resampling because it is a closed-form function of the estimated score skewness and kurtosis. Simulations show that our proposal can make a difference in size control with as few as 10 clusters. Keywords: Cluster robust inference, Cramér-Edgeworth expansion, Asymptotic refinement
Read the full abstract

Combinatorics

Coloopless zonotopes and counterexamples to the Shifted Lonely Runner Conjecture

Henze and Malikiosis (2017) have shown that the Lonely Runner Conjecture (LRC) can be restated as a convex-geometric question on the so-called LR zonotopes, lattice zonotopes with one more generator than their dimension. This relation naturally suggests a more general statement, the shifted LRC, the zonotopal version of which concerns a classical parameter, the covering radius. In this paper we do two things: 1) We show explicit counterexamples to both the shifted Lonely Runner Conjecture (starting at $n=5$) and to the Lonely Vector Property of Malikiosis-Schymura-Santos (2025). 2) We push the analogies between the two versions of LRC and their zonotopal counterparts, in particular highlighting that the proofs of the finite-checking Theorems A and B in Malikiosis-Schymura-Santos (2025) are more transparent, and the statements more general, if regarded in terms of two quite general classes of lattice zonotopes: the coloopless zonotopes that we introduce here and the cosimple ones, already defined by them.
Read the full abstract

Statistics Theory

Binary Expansion Group Intersection Network

Conditional independence is central to modern statistics, but beyond special parametric families it rarely admits an exact covariance characterization. We introduce the binary expansion group intersection network (BEGIN), a distribution-free graphical representation for multivariate binary data and bit-encoded multinomial variables. For arbitrary binary random vectors and bit representations of multinomial variables, we prove that conditional independence is equivalent to a sparse linear representation of conditional expectations, to a block factorization of the corresponding interaction covariance matrix, and to block diagonality of an associated generalized Schur complement. The resulting graph is indexed by the intersection of multiplicative groups of binary interactions, yielding an analogue of Gaussian graphical modeling beyond the Gaussian setting. This viewpoint treats data bits as atoms and local BEGIN molecules as building blocks for large Markov random fields. We also show how dyadic bit representations allow BEGIN to approximate conditional independence for general random vectors under mild regularity conditions. A key technical device is the Hadamard prism, a linear map that links interaction covariances to group structure.
Read the full abstract

Analysis of PDEs

Some remarks on patterns for semilinear Neumann problems

We study semilinear elliptic equations \begin{equation*} \begin{cases} -\Delta u = f(u) & \text{in } \Omega, \\ \partial_\nu u = 0 & \text{on } \partial\Omega, \end{cases} \end{equation*} with homogeneous Neumann boundary conditions in bounded domains. A classical result by Casten-Holland and Matano shows that stable nonconstant solutions cannot exist in convex domains, although unstable spatial patterns may still occur. In this paper we investigate rigidity properties of classical solutions without imposing stability assumptions and aim to identify structural conditions on the nonlinearity ensuring that all solutions are constant. We prove that every classical solution of the Neumann problem is constant provided the nonlinearity satisfies a suitable `monotonicity' condition, which includes the cases where the nonlinearity has a fixed sign or changes sign in a controlled way around one of its zeros. This yields a rigidity result depending solely on the structure of the nonlinearity and does not require convexity assumptions on the domain. We also discuss the sharpness of our assumptions by constructing examples of nonlinearities for which nonconstant solutions exist. In particular, inspired by the approach of Lin-Ni-Takagi, we consider exponential-type nonlinearities in dimension $N=2$, and show that when a parameter crosses a critical threshold, the associated Neumann problem admits nontrivial and nonconstant solutions for sufficiently small diffusion.
Read the full abstract

Optimization and Control

Nested Extremum Seeking Converges to Stackelberg Equilibrium

The nested Extremum Seeking (nES) algorithm is a model-free optimization method that has been shown to converge to a neighborhood of a Nash equilibrium. In this work, we demonstrate that the same nES dynamics can instead be made to converge to a neighborhood of a Stackelberg (leader–follower) equilibrium by imposing a different scaling law on the algorithm's design parameters. For the two–level nested case, using Lie–bracket averaging and singular perturbation arguments, we provide a rigorous stability proof showing semi-global practical asymptotic convergence to a Stackelberg equilibrium under appropriate time-scale separation. The results reveal that equilibrium selection, Nash versus Stackelberg, depends not on modifying the closed-loop dynamics, but on the hierarchical scaling of design parameters and the induced time-scale structure. We demonstrate this effect using a simple quadratic example and the canonical Fish War game. The Stackelberg variant of nES provides a model-free framework for hierarchical optimization in multi-time-scale systems, with potential applications in power grids, networked dynamical systems, and tuning of particle accelerators.
Read the full abstract

All authors Brad Ratto, Alan Williams, Miroslav Krstić, Tamer Başar, Alexander Scheinker

Dynamical Systems

Distributed MPC For Coordinated Path-Following

In this paper, we consider a distributed model predictive control (MPC) algorithm for coordinated path-following. Relying on the time-critical cooperative path-following framework, which decouples space and time and reduces the coordination problem to a one-dimensional setting, we formulate a distributed MPC scheme for time coordination. Leveraging properties of the normalized Laplacian, we decouple the MPC dynamics into independent modes and derive a recursive relation linking current and predicted states. Using this structure, we prove that, for prediction horizon $K=1$ and a fixed connected communication network, the system is exponentially stable even in the presence of path-following errors. This provides a first result on the convergence analysis of discrete-time distributed MPC schemes within this framework. The proposed approach enables scalable and efficient real-time implementation with low communication overhead. Moreover, in contrast to the time-critical cooperative path-following framework, the optimization-based structure relaxes the reliance on preplanning by allowing the incorporation of mission-specific requirements, such as vehicle limitations, collision avoidance, and conflict resolution. Simulation results demonstrate applicability to complex scenarios, highlighting agility and exponential convergence under communication failures.
Read the full abstract

All authors Lusine Poghosyan, Anna Manucharyan, Mikayel Aramyan, Naira Hovakimyan, Tigran Bakaryan

Group Theory

Splitting of Clifford groups associated to finite abelian groups

The Clifford group associated with a finite abelian group fits into an extension of the symplectic group by the underlying phase space. We prove that this extension splits as a semidirect product if and only if the order of the underlying abelian group is not divisible by four. The obstruction to splitting is controlled entirely by the 2-primary component and vanishes precisely when this component is trivial or cyclic of order two. This confirms a conjecture of Korbelář and Tolar, extending their cyclic result to arbitrary finite abelian groups.
Read the full abstract

Analysis of PDEs

Low-regularity global well-posedness theory for the generalized Zakharov-Kuznetsov equation on $\mathbb{R} \times \mathbb{T}$ and polynomial growth of higher Sobolev norms

We address the Cauchy problem for the $k$-generalized Zakharov-Kuznetsov equation ($k$-gZK) posed on $\mathbb{R}^2$ and on $\mathbb{R} \times \mathbb{T}$. By applying established and recently developed linear and bilinear Strichartz-type estimates within the framework of the $I$-method, we obtain the following results: $\bullet$ The Zakharov-Kuznetsov equation is globally well-posed in $H^s(\mathbb{R} \times \mathbb{T})$ for every $s>\frac{11}{13}$. $\bullet$ The modified Zakharov-Kuznetsov equation is globally well-posed in $H^s(\mathbb{R}^2)$ for every $s>\frac{2}{3}$ and in $H^s(\mathbb{R} \times \mathbb{T})$ for every $s>\frac{36}{49}$. Moreover, we show that the $H^s(\mathbb{R} \times \mathbb{T})$-norm of smooth global real-valued solutions of $k$-gZK grows at most polynomially in time for every $k\geq 1$.
Read the full abstract

Optimization and Control · Cross-listed from econ.TH

Adversarial Selection

In many institutional settings, $k$ items are selected with the goal of representing the underlying distribution of claims, opinions, or characteristics in a large population. We study environments with two adversarial parties whose preferences over the selected items are commonly known and opposed. We propose the Quantile Mechanism: one party partitions the population into $k$ disjoint subsets, and the other selects one item from each subset. We show that this procedure is optimally representative among all feasible mechanisms, and illustrate its use in jury selection, multi-district litigation, and committee formation.
Read the full abstract

All authors Alma Cohen, Alon Klement, Zvika Neeman, Eilon Solan

Logic

Linear theories of global fields with absolute values

We study the theory of a global field k as a k-vector space with a predicate for one of the absolute values on k. For example, we prove that in this language a global field with an ultrametric or real archimedean absolute value has a decidable theory, while with a complex absolute value the theory is always undecidable. We also study the existential theories and axiomatize k together with predicates for all non-complex absolute values on k simultaneously.
Read the full abstract

Combinatorics

Hamilton decompositions of the directed 3-torus: a return-map and odometer view

We prove that the directed 3-torus D_3(m), or equivalently the Cartesian product of three directed m-cycles, admits a decomposition into three arc-disjoint directed Hamilton cycles for every integer m >= 3. The proof reduces Hamiltonicity to the m-step return maps on the layer section S=i+j+k=0. For odd m, five Kempe swaps of the canonical coloring produce return maps that are explicitly affine-conjugate to the standard 2-dimensional odometer. For even m, a sign-product invariant rules out Kempe-from-canonical constructions, and a different low-layer witness reduces after one further first-return map to a finite-defect clock-and-carry system. The remaining closure is a finite splice analysis, and the case m=4 is handled separately by a finite witness. A Lean 4 formalization accompanies the construction.
Read the full abstract

Numerical Analysis

A note on superconvergence in projection-based numerical approximations of eigenvalue problems for Fredholm integral operators

This paper studies the eigenvalue problem $K \psi = \lambda \psi$ associated with a Fredholm integral operator $K$ defined by a smooth kernel. The focus is on analyzing the convergence behaviour of numerical approximations to eigenvalues and their corresponding spectral subspaces. The interpolatory projection methods are employed on spaces of piecewise polynomials of even degree, using $2r+1$ collocation points that are not restricted to Gauss nodes. Explicit convergence rates are established, and the modified collocation method attains faster convergence of approximation of eigenvalues and associated eigenfunctions than the classical collocation scheme. Moreover, it is shown that the iteration yields superconvergent approximations of eigenfunctions. Numerical experiments are presented to validate the theoretical findings.
Read the full abstract

Group Theory

Coarse separation and splittings in right-angled Artin groups

In this article, we characterise geometrically when a right-angled Artin group splits over an abelian subgroup. More precisely, given a finite graph $\Gamma$, we show that $A(\Gamma)$ splits over an abelian subgroup if and only if it is coarsely separable by a family of subexponential growth, which amounts to saying that $\Gamma$ is complete or separated by a complete subgraph.
Read the full abstract

Information Theory

Quadratic Residue Codes over $\mathbb{Z}_{121}$

In this paper, we construct a special family of cyclic codes, known as quadratic residue codes of prime length \( p \equiv \pm 1 \pmod{44} ,\) \( p \equiv \pm 5 \pmod{44} ,\) \( p \equiv \pm 7 \pmod{44} ,\) \( p \equiv \pm 9 \pmod{44} \) and \( p \equiv \pm 19 \pmod{44} \) over $\mathbb{Z}_{121}$ by defining them using their generating idempotents. Furthermore, the properties of these codes and extended quadratic residue codes over $\mathbb{Z}_{121}$ are discussed, followed by their Gray images. Also, we show that the extended quadratic residue code over $\mathbb{Z}_{121}$ possesses a large permutation automorphism group generated by shifts, multipliers, and inversion, making permutation decoding feasible. As examples, we construct new codes with parameters $[55,5,33]$ and $[77,7,44].$
Read the full abstract

Group Theory

Abstract twisted Brin–Thompson groups

Given a group $G$ acting faithfully on a set $S$, one gets a simple group denoted $SV_G$, called a twisted Brin–Thompson group. In this paper we drop the faithfulness assumption, and get an abstract version of a twisted Brin–Thompson group $SV_G$. While the resulting group is not simple, since $SV_G$ surjects onto $SV_{G/\ker(G \curvearrowright S)}$, we prove that every proper normal subgroup of $SV_G$ lies in the kernel of this surjection, so $SV_G$ is “relatively simple”. The advantage is that now we can prove that every finitely presented simple group embeds in a finitely presented abstract twisted Brin–Thompson group intersecting this kernel trivially. In particular, if the Boone–Higman conjecture is true, then so is a related conjectural characterization of groups with solvable word problem, arising purely in the world of twisted Brin–Thompson groups. We also prove a variety of additional results about abstract twisted Brin–Thompson groups, some of which are new even in the faithful case: they are all uniformly perfect, have property NL and property FW$_\infty$, are boundedly acyclic and $\ell^2$-invisible, and are $C^*$-simple as soon as they have trivial amenable radical. Along the way we formulate a new general criterion for $\ell^2$-invisibility that is interesting in its own right.
Read the full abstract

Algebraic Geometry

Bounds on the Mordell-Weil rank of elliptic fibrations

We prove that the Mordell-Weil group of a higher dimensional elliptic fibration naturally embeds into that of a suitable elliptic surface. We give sufficient conditions for the existence of such surfaces. We apply our result to obtain explicit and uniform bounds for the Mordell-Weil rank of elliptic threefolds of Kodaira dimension zero, including Calabi-Yau threefolds, confirming predictions from physics. We prove new explicit bounds for a broad class of elliptic fourfolds. These results suggest a general linear bound for the Mordell-Weil rank in terms of the dimension of the elliptic fibration, which we formulate as a conjecture.
Read the full abstract

All authors Antonella Grassi, Rick Miranda, Kapil Paranjape, Vasudevan Srinivas, Timo Weigand

Mathematical Physics · Cross-listed from hep-th

Duality-Invariant Higher-Derivative Corrections to Charged Stringy Black Holes

We study duality-invariant higher-derivative corrections to the charged black hole geometry in two-dimensional heterotic string theory. We illustrate how the conventional perturbative approach to determine the corrected geometry breaks down. Using a non-perturbative (in $\alpha'$) parametrization of the solution, we find the corrected charge-to-mass ratio for extremal black holes. We remark on the results in relation to the weak gravity conjecture. We also consider the entropy of the extremal black hole within the attractor mechanism and find that the two-derivative entropy is not renormalized to any order. We make comments on interpretations of the results and their extension to near-extremal black holes.
Read the full abstract

Number Theory

2- and 3-Dissections of Second-, Sixth-, and Eighth-Order Mock Theta Functions

In this paper, we develop a systematic method for obtaining and proving $m$-dissections of mock theta functions. In 2014, Hickerson and Mortenson showed how to derive and prove identities for Ramanujan's mock theta functions and Hecke-type indefinite theta series using Appell–Lerch sums. We build on their transformation formula method, combining it with symbolic computations and algorithms for the theory of modular functions. We focus exclusively on the cases of 2- and 3-dissections.
Read the full abstract

Numerical Analysis · Cross-listed from cs.LG

Learning Mesh-Free Discrete Differential Operators with Self-Supervised Graph Neural Networks

Mesh-free numerical methods provide flexible discretisations for complex geometries; however, classical meshless discrete differential operators typically trade low computational cost for limited accuracy or high accuracy for substantial per-stencil computation. We introduce a parametrised framework for learning mesh-free discrete differential operators using a graph neural network trained via polynomial moment constraints derived from truncated Taylor expansions. The model maps local stencils relative positions directly to discrete operator weights. The current work demonstrates that neural networks can learn classical polynomial consistency while retaining robustness to irregular neighbourhood geometry. The learned operators depend only on local geometry, are resolution-agnostic, and can be reused across particle configurations and governing equations. We evaluate the framework using standard numerical analysis diagnostics, showing improved accuracy over Smoothed Particle Hydrodynamics, and a favourable accuracy-cost trade-off relative to a representative high-order consistent mesh-free method in the moderate-accuracy regime. Applicability is demonstrated by solving the weakly compressible Navier-Stokes equations using the learned operators.
Read the full abstract

All authors Lucas Gerken Starepravo, Georgios Fourtakas, Steven Lind, Ajay B. Harish, Tianning Tang, Jack R. C. King

Probability · Cross-listed from q-fin.RM

Ordering results for extreme claim amounts based on random number of claims

Consider two sequences of heterogeneous and independent portfolios of risks $T_1,T_2,\ldots$ and $T^*_{1}, T^*_{2},\ldots$ and, let $N_1$ and $N_2$ be two positive integer-valued random variables, independent of $T_i'$ and $T^*_i$, respectively. In this article, we investigate different stochastic inequalities involving $\min\{T_1,\ldots,T_{N_1}\}$ and $\min\{T^*_1,\ldots,T^*_{N_2}\},$ and $\max\{T_1,\ldots,T_{N_1}\}$ and $\max\{T^*_1,\ldots,T^*_{N_2}\}$ in the sense of usual stochastic order and reversed hazard rate order concerning maltivariate chain majorization order. These new results strengthen and generalize some of the well known results in the literature, including \cite{barmalzan2017ordering}, \cite{balakrishnan2018} and \cite{kundu2021_shock} for the case of random claim sizes. Different numerical examples are provided to highlight the applicability of this work. Finally, some interesting applications of our results in reliability theory and auction theory are presented.
Read the full abstract

Group Theory

On the multiplicative group of a two-sided skew brace of solvable type

We prove that if $(B,+,\cdot)$ is a two-sided skew brace whose additive group is solvable, then every finite quotient of the multiplicative group $(B,\cdot)$ is solvable. In particular, our result recovers Nasybullov's theorem in the finite case ~\cite[Theorem~4.3(1)]{Nas} and extends it to arbitrary two-sided skew braces of solvable type.
Read the full abstract

Differential Geometry

A comment on full discretized isothermic tori in Euclidean spaces

Using discretized orthogonal systems (curvature line systems) with periodicity, created using Darboux transformations and their permutability, we have discrete and semi-discrete k-dimensional isothermic tori which are full in n-dimensional Euclidean space, for any natural numbers k between 2 nd n.
Read the full abstract

Combinatorics

Two-color Ramsey lower bounds for bounded degree hypergraphs

We consider Ramsey numbers of bounded-degree uniform hypergraphs. In particular, we prove that for every $k\ge3$, there exists a constant $c_k>0$ such that, for all sufficiently large $\Delta$ and every $n\ge2^\Delta$, there is a $k$-uniform $n$-vertex hypergraph $H$ with maximum degree at most $\Delta$ satisfying \[ r(H)\ge \tw_{k-1}\!\bigl(c_k\Delta\log\log\Delta\bigr)\,n. \] Here $\tw_j$ denotes the tower function of height $j$. This constitutes the first progress towards a problem posed by Conlon, Fox and Sudakov.
Read the full abstract

Numerical Analysis · Cross-listed from cs.LG

Polynomial Speedup in Diffusion Models with the Multilevel Euler-Maruyama Method

We introduce the Multilevel Euler-Maruyama (ML-EM) method compute solutions of SDEs and ODEs using a range of approximators $f^1,\dots,f^k$ to the drift $f$ with increasing accuracy and computational cost, only requiring a few evaluations of the most accurate $f^k$ and many evaluations of the less costly $f^1,\dots,f^{k-1}$. If the drift lies in the so-called Harder than Monte Carlo (HTMC) regime, i.e. it requires $\epsilon^{-\gamma}$ compute to be $\epsilon$-approximated for some $\gamma>2$, then ML-EM $\epsilon$-approximates the solution of the SDE with $\epsilon^{-\gamma}$ compute, improving over the traditional EM rate of $\epsilon^{-\gamma-1}$. In other terms it allows us to solve the SDE at the same cost as a single evaluation of the drift. In the context of diffusion models, the different levels $f^{1},\dots,f^{k}$ are obtained by training UNets of increasing sizes, and ML-EM allows us to perform sampling with the equivalent of a single evaluation of the largest UNet. Our numerical experiments confirm our theory: we obtain up to fourfold speedups for image generation on the CelebA dataset downscaled to 64x64, where we measure a $\gamma\approx2.5$. Given that this is a polynomial speedup, we expect even stronger speedups in practical applications which involve orders of magnitude larger networks.
Read the full abstract