Search arXiv⌕ Search

arXiv subjects

Search papers

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

At least 1,639 records · Page 91Linked to original sources

A direct proof of the existence of MME for finite horizon Sinai billiards

The Sinai billiard map $T$ on the two-torus, i.e., the periodic Lorentz gas, is a discontinuous map. We provide two separate conditions for a finite horizon billiard to admit equilibrium states. The first one -- namely \emph{negligible singularities} -- only concerns the billiard table and implies upper semi-continuity for the metric pressure map of $T$. The condition of negligible singularities is shown to be generic (dense $G_δ$) among billiard tables with $C^{3+α}$ boundaries, in the $C^{3+α}$ topology, and implies the sparse recurrence condition for the null potential. Compactness of the set of $T$-invariant measures is proved. For finite horizon Sinai billiards, we provide a bound on the defect of upper semi-continuity of the metric pressure map. From it, a condition of existence of equilibrium states only involving potentials is deduced. In both situations, for the potentials $g \equiv 0$ and $g = -h_{\rm top}(Φ^1) τ$, we recover and extend the recent results of the existence of measures of maximal entropy (MME) for both the map $T$ and the flow $Φ^t$, due to Baladi and Demers for $T$, jointly with Carrand for $Φ^t$. Upper semi-continuity of the metric entropy is shown. Assuming negligible singularities, equidistribution of the periodic orbits with respect to the unique MME is proved. A bound on the topological tail entropy is given.

math.DS↗

Are EEG Sequences Time Series? EEG Classification with Time Series Models and Joint Subject Training

As with most other data domains, EEG data analysis relies on rich domain-specific preprocessing. Beyond such preprocessing, machine learners would hope to deal with such data as with any other time series data. For EEG classification many models have been developed with layer types and architectures we typically do not see in time series classification. Furthermore, typically separate models for each individual subject are learned, not one model for all of them. In this paper, we systematically study the differences between EEG classification models and generic time series classification models. We describe three different model setups to deal with EEG data from different subjects, subject-specific models (most EEG literature), subject-agnostic models and subject-conditional models. In experiments on three datasets, we demonstrate that off-the-shelf time series classification models trained per subject perform close to EEG classification models, but that do not quite reach the performance of domain-specific modeling. Additionally, we combine time-series models with subject embeddings to train one joint subject-conditional classifier on all subjects. The resulting models are competitive with dedicated EEG models in 2 out of 3 datasets, even outperforming all EEG methods on one of them.

cs.LG↗

Enhanced Whispering Gallery Mode Phase Shift using Indistinguishable Photon Pairs

We present a theoretical investigation of a whispering gallery mode (WGM) resonator coupled to a Mach-Zehnder interferometer (MZI) and show a bimodal coincidence transmission spectrum when the input state is an indistinguishable photon pair. This is due to the doubled WGM phase shift experienced by the path-entangled state in the interferometer. Further, we model the noise in a WGM resonance shift measurement comparing photon pairs with a coherent state. At least a four-fold improvement in the signal-to-noise ratio (SNR) is possible, with clear implications for quantum-enhanced WGM sensing.

quant-ph↗

Voting with Partial Orders: The Plurality and Anti-Plurality Classes

In the theory of voting, the Plurality rule for preferences that come in the form of linear orders selects the alternatives most frequently appearing in the first position of those orders, while the Anti-Plurality rule selects the alternatives least often occurring in the final position. We explore extensions of these rules to preferences that are partial orders, offering axiomatic characterisations for them.

cs.GT↗

SO(5) multicriticality in two-dimensional quantum magnets

We resolve the nature of the quantum phase transition between antiferromagnetic and valence-bond solid ground states in square-lattice spin-1/2 magnets - the putative deconfined quantum-critical point. Using quantum Monte Carlo simulations of J-Q models, in which Heisenberg exchange J competes with multi-spin interactions Qn formed by n singlet projectors, we demonstrate first-order discontinuities increasing with n. For n=2 and n=3, the first-order signatures are very weak and we extract universal scaling dimensions of a proximate multicritical point. Their values coincide closely with known bounds for an SO(5) symmetric conformal field theory (CFT) containing a relevant SO(5) singlet operator that destabilizes the generic deconfined critical point. We characterize the emergent SO(5) symmetry by its scaling dimensions of the leading irrelevant perturbations. Using the J-Q2-Q6 model, we tune the transition from weak to stronger first-order and find scaling of the coexisting order parameters with a large exponent beta ~ 0.85, in agreement with the relevant scaling dimensions of the multicritical point. The exponent that had previously been identified with the correlation length does not have a corresponding CFT level, and we explain it as pseudo-critical length scale of bubbles of one phase inside the other that are compressed in size by an SO(5)-violating irrelevant repulsive interaction between the two order parameters. The bubble scale reflects a hierarchy of excitation gaps generated by the SO(5) perturbation in the neighborhood of the first-order transition, akin to a Higgs mechanism. We argue that the multicritical point is the tip of a gapless spin liquid phase in frustrated Heisenberg models into which the J-Q models can be deformed, but the spin liquid, and perhaps the multicritical point itself, exist outside the regime amenable to sign-free simulations.

cond-mat.str-el↗

Minimal finite models of wedges of spheres with at most eight points

A minimal finite model of a topological space is a finite $T_0$-space of least cardinality that is weak homotopy equivalent to it. We determine, up to homeomorphism, the minimal finite models of eight wedges of one- and two-dimensional spheres. The spaces $S^{2}\vee S^{1}$ and $S^{2}\vee S^{2}$ have exactly two and three minimal finite models, respectively, all with seven points, whereas $S^{1}\vee S^{1}\vee S^{2}$ has exactly seven minimal finite models, $S^{1}\vee S^{1}\vee S^{1}\vee S^{2}$ has one, $S^{1}\vee S^{2}\vee S^{2}$ has six, $S^{2}\vee S^{2}\vee S^{2}$ has five, $S^{2}\vee S^{2}\vee S^{2}\vee S^{2}$ has three, and $S^{1}\vee S^{1}\vee S^{2}\vee S^{2}$ has two, all with eight points. To prove this we classify the connected finite $T_0$-spaces of height $2$ without beat points that have seven or eight points, and we identify their weak homotopy types by means of McCord's theorem and explicit collapse and suspension arguments for their two-dimensional order complexes. Together with the known results for spheres and wedges of circles, this determines, up to weak homotopy equivalence, all path-connected spaces whose minimal finite models have at most eight points.

math.AT↗

Mean-field limit of non-exchangeable multi-agent systems over hypergraphs with unbounded rank

Interacting particle systems are known for their ability to generate large-scale self-organized structures from simple local interaction rules between each agent and its neighbors. In addition to studying their emergent behavior, a main focus of the mathematical community has been concentrated on deriving their large-population limit. In particular, the mean-field limit consists of describing the limit system by its population density in the product space of positions and labels. The strategy to derive such limits is often based on a careful combination of methods ranging from analysis of PDEs and stochastic analysis, to kinetic equations and graph theory. In this article, we focus on a generalization of multi-agent systems that includes higher-order interactions, which has largely captured the attention of the applied community in the last years. In such models, interactions between individuals are no longer assumed to be binary (i.e. between a pair of particles). Instead, individuals are allowed to interact by groups so that a full group jointly generates a non-linear force on any given individual. The underlying graph of connections is then replaced by a hypergraph, which we assume to be dense, but possibly non uniform and of unbounded rank. For the first time in the literature, we show that when the interaction kernels are regular enough, then the mean-field limit is determined by a limiting Vlasov-type equation, where the hypergraph limit is encoded by a so-called UR-hypergraphon (unbounded-rank hypergraphon), and where the resulting mean-field force admits infinitely-many orders of interactions.

math.AP↗

Noncontextuality inequalities for prepare-transform-measure scenarios

We provide the first systematic technique for deriving witnesses of contextuality in prepare-transform-measure scenarios. More specifically, we show how linear quantifier elimination can be used to compute a polytope of correlations consistent with generalized noncontextuality in such scenarios. This polytope is specified as a set of noncontextuality inequalities that are necessary and sufficient conditions for observed data in the scenario to admit of a classical explanation relative to any linear operational identities, if one ignores some constraints from diagram preservation. While including these latter constraints generally leads to tighter inequalities, it seems that nonlinear quantifier elimination would be required to systematically include them. We also provide a linear program which can certify the nonclassicality of a set of numerical data arising in a prepare-transform-measure experiment. We apply our results to get a robust noncontextuality inequality for transformations that can be violated within the stabilizer subtheory. Finally, we give a simple algorithm for computing all the linear operational identities holding among a given set of states, of transformations, or of measurements.

quant-ph↗

Simulation in discrete choice models evaluation: SDCM, a simulation tool for performance evaluation of DCMs

Discrete choice models are widely used in economics, transportation, and marketing to explain and predict how decision makers choose among a finite set of alternatives. The resulting choices depend on the socioeconomic characteristics of the decision makers, the attributes of the alternatives, and the conditions under which the decision is made. Because these factors interact in complex ways, and because real experiments raise confidentiality concerns, take time, and are costly, controlled experimentation with real respondents is often impractical. Simulation offers an alternative in which the data-generating process, and hence the true parameters, are known. This paper connects discrete choice modeling, experimental design, and simulation. We review the three areas in a common notation and state the results on which the simulation approach relies. We then formalize a simulation tool that combines an experimental design with a behavioral model of artificial agents to generate synthetic stated choice data, and we prove that the generated choices follow a panel mixed multinomial logit model with known parameters. On this basis we define measures of estimator performance and of the discrepancy between simulated and real data, illustrate the tool with a stated choice experiment on the willingness to pay for environmental attributes of roses, and discuss the behavioral factors that separate simulated choices from real ones.

econ.GN↗

Opinion dynamics on switching networks

We study opinion dynamics over a two-layer multiplex network with weighted, directed influence and reciprocal links. Within each layer, an agent weighs its neighbors equally, so the influence of a neighbor is inversely proportional to the agent's number of neighbors in that layer. Agents update their opinions over time to coordinate with their neighbors. However, the frequency of agents' interactions with neighbors in different network layers differs: one layer is always active, whereas the other one is active only once every $k$ time steps. Consequently, the multiplex network's adjacency matrices are time-varying. We characterize how the activation frequency of the second layer impacts the convergence of the opinion dynamics process. Without requiring agents to place positive weight on their own opinions, we prove that, for every activation period $k$, the opinions reach a consensus whose value depends on $k$, and that the asymptotic per-step convergence factor equals the $k$th root of the second-largest eigenvalue modulus of the one-period transition matrix. As $k$ grows, this factor generically converges to that of the always-active layer when this layer is connected; when it is disconnected, the time to consensus generically grows linearly in $k$. Numerical experiments on 1,000 random multiplex networks confirm these results.

physics.soc-ph↗

Monotonicity properties of Hadamard weighted geometric means of bounded sets of positive kernel operators

We prove new monotonicity properties for joint and generalized spectral radius and their essential versions, operator norm, Hausdorff measure of non-compactness and numerical radius of products and sums of Hadamard weighted geometric means of positive kernel operators on appropriate Banach function spaces. Our results extend recent results on weighted geometric symmetrizations of bounded sets of such operators defined on $L^2$. To our knowledge, several proved properties in this generality are new even in the finite dimensional case and also in the case of singelton sets.

math.FA↗

A multi-objective mixed integer linear programming model for supply chain planning of 3D printing

Three-dimensional (3D) printing makes customized parts in small batches, but 3D printers are expensive and slow, so the grouping, orientation, and sequencing of parts decide both delivery times and printer use. We study the allocation of parts with due dates to jobs on identical powder bed 3D printers, where the orientation of a part sets its height and the plate area it occupies. We formulate a bi-objective mixed integer linear programming model that minimizes the weighted earliness and tardiness of the parts and the unused plate area of the jobs, and we show that the optimal timing of a fixed job sequence is an isotonic regression. The Pareto front is computed with the epsilon-constraint method, and a non-dominated sorting genetic algorithm built on this timing result handles larger instances. On random instances, choosing the orientation lowers the earliness and tardiness by 63\% on average compared with the orientation given in the data. For 15 to 40 parts, the genetic algorithm keeps 98\% to 99\% of the hypervolume of the best known fronts in less than a minute per run on average, whereas a time-limited epsilon-constraint method keeps about half of it with 30 and 40 parts.

math.OC↗

Guard Analysis and Safe Erasure Gradual Typing: a Type System for Elixir

We formalize a new type system for Elixir, a dynamically typed functional programming language of growing popularity that runs on the Erlang virtual machine. Our system combines gradual typing with semantic subtyping to enable precise, sound, and practical static type analysis, without requiring any changes to Elixir's compilation pipeline or runtime. Type soundness is ensured by leveraging runtime checks -- both implicit, from the Erlang VM, and explicit, via developer-written guards. Central to our approach are two key innovations: the notion of "strong functions", which can be assigned precise types even when applied to inputs that may fall outside their intended domain; and a fine-grained analysis of guards that enables accurate type refinement for case expressions and guarded function definitions. While type information is erased before execution and not used by the compiler, our "safe erasure" gradual typing strategy maintains soundness and expressiveness without compromising compatibility or performance. This work lays the theoretical foundation for Elixir's new type system, outlines its integration into recent versions of the language, and demonstrates its effectiveness on large-scale industrial codebases.

cs.PL↗

Tandem jets in crossflow: a one-parameter dilution law for the trajectory and the velocity decay

The front jet of a tandem pair in crossflow is not passive. This paper treats the mean-flow phenomenology and its scaling; the mechanism is left to the companion paper. In 40\ kHz time-resolved PIV measurements covering nine spacing--velocity-ratio conditions ($L=2d$, $3d$, $4d$; $r\approx3$, $5$, $7$) and a single-jet baseline, the front-jet $rd$-normalized trajectory lies below the single-jet reference in every condition (near-field segment up to 40\% shorter), while the shielded rear jet rises above the reference in eight of the nine and penetrates $1.4$--$2.0$ times deeper than the front jet. The mean pressure and reverse-flow statistics are inconsistent with the two classical explanations: the inter-jet region is pressurized rather than sub-atmospheric, the lateral pressure difference across the front jet is weakened rather than enhanced (within reconstruction uncertainty), and the reverse-flow region does not follow the ordering of the modification. In natural coordinates the centerline decay re-scales onto the single-jet law at the effective velocity ratio (seven of nine within 18\%), so bending and decay are two readouts of one dilution, which whole-trajectory matching condenses into the one-parameter law $r_{\mathrm{eff}}/r=r/(r+c)$, $c=3.31$, with the velocity ratio as the primary control and the spacing entering through the near-field capacity ($c_h=2.50$--$3.30$). The law has one fitted constant and needs only the single-jet reference family; leave-one-condition-out predictions reproduce each held-out condition within 15\%, so the procedure is portable to any tandem array in which the gap recirculation closes.

physics.flu-dyn↗

Risk-Aware Alignment Must Incorporate Reward Uncertainty

We consider the problem of policy optimization with noisy rewards, where the policy is constrained to be in the neighborhood of a reference policy.We study the risk of the policy optimization process, defined to be the chance of obtaining a policy worse off than the reference policy. In an idealized multivariate Gaussian model for the noisy reward function, we show that the regularizer achieving optimal risk in a suitably local neighborhood of the reference policy is the quadratic norm weighted by the reward covariance matrix. This result is used to develop a new policy optimization algorithm based on regularization via reward variance estimates. Experiments across synthetic bandit models, large language models and reward models show that this algorithm yields more robust alignment outcomes than existing approaches.

cs.AI↗

Spherical growth of reciprocal classes in the Hecke Groups

Let $r \in \mathbb{Z}$, $r>2$, and let $Γ_{2r} \cong (\mathbb{Z}/2\mathbb{Z}) \ast (\mathbb{Z}/2r\mathbb{Z})$ be the Hecke group. Let $\mathcal{N}_\ell$ denote the set of conjugacy classes of reciprocal elements with cyclically reduced word length exactly $\ell$ in the standard generating set of $Γ_{2r}$. Let $P(x)$ be the polynomial \[ P(x) = x^{r+1} - 2\sum_{i=1}^{r-1} x^{i} - 1. \] Let $ρ$ be the unique positive root of $P$ and $C$ be a constant depending on $r.$ We prove that, as $\ell \to \infty$, \[|\mathcal{N}_\ell| \sim Cρ^{\left\lfloor \tfrac{\ell+1}{2} \right\rfloor}. \] Our approach exploits the free product structure of $Γ_{2r}$, together with a combinatorial counting function and recurrence relations arising from cyclically reduced representatives. We further show that the asymptotic growth rate of primitive reciprocal conjugacy classes of word length $\ell$ coincides with that of $\mathcal{N}_\ell$. This result extends previous work for odd values of $p$ and provides a uniform asymptotic bound for all Hecke groups.

math.GR↗

Relative Trace Formula And Simultaneous Nonvanishing for GL_3 x GL_2 and GL_3 x GL_1 L-functions

Fix a Dirichlet character $χ$ and a cuspidal GL$(2)$ eigenform $ϕ$ with relatively prime conductors. Then we show that there are infinitely many cusp forms $π$ on GL$(3)$ such that $L(1/2, π\times χ)$ and $L(1/2, π\times ϕ)$ are simultaneously non-zero. We achieve this by use of Jacquet's Relative Trace Formula. We derive an expression of the average over the GL$(3)$ cuspidal spectrum as a sum of a non-zero main term and two subsidiary terms which are forced to be zero for large enough level by use of a suitable test function. This modest article is dedicated to the memory of Harish Chandra, on the occasion of his hundredth birthday.

math.NT↗