Search

The task of Visual Place Recognition (VPR) is to predict the location of a query image from a database of geo-tagged images. Recent studies in VPR have highlighted the significant advantage of employing pre-trained foundation models like DINOv2 for the VPR task. However, these models are often deemed inadequate for VPR without further fine-tuning on task-specific data. In this paper, we propose a simple yet powerful approach to better exploit the potential of a foundation model for VPR. We first demonstrate that features extracted from self-attention layers can serve as a powerful re-ranker for VPR. Utilizing these features in a zero-shot manner, our method surpasses previous zero-shot methods and achieves competitive results compared to supervised methods across multiple datasets. Subsequently, we demonstrate that a single-stage method leveraging internal ViT layers for pooling can generate global features that achieve state-of-the-art results, even when reduced to a dimensionality as low as 128D. Nevertheless, incorporating our local foundation features for re-ranking, expands this gap. Our approach further demonstrates remarkable robustness and generalization, achieving state-of-the-art results, with a significant gap, in challenging scenarios, involving occlusion, day-night variations, and seasonal changes.

We use the numerical continued fraction method to investigate quasinormal mode spectra of extremal and non-extremal Reissner-Nordstr\"om black holes in the low and intermediate damping regions. In the extremal case, we develop techniques that significantly expand the calculated spectrum from what had previously appeared in the literature. This allows us to determine the asymptotic behavior of the extremal spectrum in the high damping limit, where there are conflicting published results. Our investigation further supports the idea that the extremal limit of the non-extremal case, where the charge approaches the mass of the black hole in natural units, leads to the same vibrational spectrum as in the extremal case despite the qualitative differences in their topology. In addition, we numerically explore the quasinormal mode spectrum for a Reissner-Nordstr\"om black hole in the small charge limit., Comment: 13 pages, 3 figures, regular article, minor corrections of the language, an acknowledgment section is added, the paper is accepted for publication in PRD

Integrated correlation functions in $\mathcal{N}=4$ supersymmetric Yang--Mills theory with gauge group $SU(N)$ can be expressed in terms of the localised $S^4$ partition function, $Z_N$, deformed by a mass $m$. Two such cases are $\mathcal{C}_N=(\text{Im} \tau)^2 \partial_\tau\partial_{\bar\tau} \partial_m^2\log Z_N\vert_{m=0}$ and $\mathcal{H}_N=\partial_m^4\log Z_N\vert_{m=0}$, which are modular invariant functions of the complex coupling $\tau$. While $\mathcal{C}_N$ was recently written in terms of a two-dimensional lattice sum for any $N$ and $\tau$, $\mathcal{H}_N$ has only been evaluated up to order $1/N^3$ in a large-$N$ expansion in terms of modular invariant functions with no known lattice sum realisation. Here we develop methods for evaluating $\mathcal{H}_N$ to any desired order in $1/N$ and finite $\tau$. We use this new data to constrain higher loop corrections to the stress tensor correlator, and give evidence for several intriguing relations between $\mathcal{H}_N$ and $\mathcal{C}_N$ to all orders in $1/N$. We also give evidence that the coefficients of the $1/N$ expansion of $\mathcal{H}_N$ can be written as lattice sums to all orders. Lastly, these large $N$ and finite $\tau$ results are used to accurately estimate the integrated correlators at finite $N$ and finite $\tau$., Comment: 30 pages plus appendices, 8 figures

Large Language Models (LLMs) are powerful tools, capable of leveraging their training on natural language to write stories, generate code, and answer questions. But can they generate functional video game levels? Game levels, with their complex functional constraints and spatial relationships in more than one dimension, are very different from the kinds of data an LLM typically sees during training. Datasets of game levels are also hard to come by, potentially taxing the abilities of these data-hungry models. We investigate the use of LLMs to generate levels for the game Sokoban, finding that LLMs are indeed capable of doing so, and that their performance scales dramatically with dataset size. We also perform preliminary experiments on controlling LLM level generators and discuss promising areas for future work.

Due to the complicated nanoscale structures of current integrated circuits(IC) builds and low error tolerance of IC image segmentation tasks, most existing automated IC image segmentation approaches require human experts for visual inspection to ensure correctness, which is one of the major bottlenecks in large-scale industrial applications. In this paper, we present the first data-driven automatic error detection approach targeting two types of IC segmentation errors: wire errors and via errors. On an IC image dataset collected from real industry, we demonstrate that, by adapting existing CNN-based approaches of image classification and image translation with additional pre-processing and post-processing techniques, we are able to achieve recall/precision of 0.92/0.93 in wire error detection and 0.96/0.90 in via error detection, respectively.

The use of supersymmetric localisation has recently led to modular covariant expressions for certain integrated correlators of half-BPS operators in $\mathcal{N} = 4$ supersymmetric Yang-Mills theory with a general classical gauge group $G_N$. Here we determine generating functions that encode such integrated correlators for any classical gauge group and provide a proof of previous conjectured formulae. This gives a systematic understanding of the relation between properties of these correlators at finite $N$ and their expansions at large $N$. In particular, it determines a duality-invariant non-perturbative completion of the large-$N$ expansion in terms of a sum of novel non-holomorphic modular functions. These functions are exponentially suppressed at large $N$ and have the form of a sum of contributions from coincident $(p, q)$-string world-sheet instantons., Comment: 25 + 15 pages; v2: minor changes, version published in JHEP

We investigate the scalar, gravitational, and electromagnetic quasinormal mode spectra of Schwarzschild anti-de Sitter black holes using the numerical continued fraction method. The spectra have similar, almost linear structures. With a few exceptions, the low overtone quasinormal modes are consistent with previously obtained results in the literature that use other numerical techniques. The intermediate and high overtone quasinormal modes, in comparison to the Schwarzschild case, converge very quickly to the asymptotic formulas previously obtained by analytic monodromy techniques. In addition, we find a connection between the analytic asymptotic formulas and the purely imaginary modes. In particular, these formulas can be used to predict the bifurcation of the lowest damped electromagnetic modes. Finally, we find no high overtone quasinormal modes with high oscillation frequency and low damping, which had been previously predicted., Comment: 21 pages, 6 figures, regular article, some typos corrected, new references added, a claim about -2i being a QNM was corrected

Ever since the Multilayered Perceptron was first introduced the connectionist community has struggled with the concept of uncertainty and how this could be represented in these types of models. This past decade has seen a lot of effort in trying to join the principled approach of probabilistic modeling with the scalable nature of deep neural networks. While the theoretical benefits of this consolidation are clear, there are also several important practical aspects of these endeavors; namely to force the models we create to represent, learn, and report uncertainty in every prediction that is made. Many of these efforts have been based on extending existing frameworks with additional structures. We present Borch, a scalable deep universal probabilistic programming language, built on top of PyTorch. The code is available for download and use in our repository https://gitlab.com/desupervised/borch.

Procedural Content Generation via Reinforcement Learning (PCGRL) foregoes the need for large human-authored data-sets and allows agents to train explicitly on functional constraints, using computable, user-defined measures of quality instead of target output. We explore the application of PCGRL to 3D domains, in which content-generation tasks naturally have greater complexity and potential pertinence to real-world applications. Here, we introduce several PCGRL tasks for the 3D domain, Minecraft (Mojang Studios, 2009). These tasks will challenge RL-based generators using affordances often found in 3D environments, such as jumping, multiple dimensional movement, and gravity. We train an agent to optimize each of these tasks to explore the capabilities of previous research in PCGRL. This agent is able to generate relatively complex and diverse levels, and generalize to random initial states and control targets. Controllability tests in the presented tasks demonstrate their utility to analyze success and failure for 3D generators., Comment: 8 pages, 9 figures

A pulse generation scheme is proposed based on a structured resonance in a cavity where the non-conventional energy distribution is concentrated in its middle part. The cavity is first used as an oscillator during the energy charging step and when a switch is activated the signal is extracted from its center. The key component of the proposed scheme is the periodic microstrip waveguide with a fourth-order degenerate band edge (DBE) of its wavenumber-frequency dispersion diagram. The DBE is an exceptional point degeneracy condition that is responsible for the energy to be localized at the cavity center and that can also have the quality factor easily destroyed by a perturbation. The waveguide is designed to have a DBE frequency of 2.86 GHz and produces pulses of approximately 0.1 V peak and 1.1 ns width., Comment: 6 pages, 7 figures

Search-based procedural content generation (PCG) is a well-known method for level generation in games. Its key advantage is that it is generic and able to satisfy functional constraints. However, due to the heavy computational costs to run these algorithms online, search-based PCG is rarely utilized for real-time generation. In this paper, we introduce mutation models, a new type of iterative level generator based on machine learning. We train a model to imitate the evolutionary process and use the trained model to generate levels. This trained model is able to modify noisy levels sequentially to create better levels without the need for a fitness function during inference. We evaluate our trained models on a 2D maze generation task. We compare several different versions of the method: training the models either at the end of evolution (normal evolution) or every 100 generations (assisted evolution) and using the model as a mutation function during evolution. Using the assisted evolution process, the final trained models are able to generate mazes with a success rate of 99% and high diversity of 86%. The trained model is many times faster than the evolutionary process it was trained on. This work opens the door to a new way of learning level generators guided by an evolutionary process, meaning automatic creation of generators with specifiable constraints and objectives that are fast enough for runtime deployment in games., Comment: 8 pages, 6 figures, and 2 tables. Published at PCGWorkshop 2022 at FDG 2022

In this paper, we present a method for automated persona-driven video game tutorial level generation. Tutorial levels are scenarios in which the player can explore and discover different rules and game mechanics. Procedural personas can guide generators to create content which encourages or discourages certain playstyle behaviors. In this system, we use procedural personas to calculate the behavioral characteristics of levels which are evolved using the quality-diversity algorithm known as Constrained MAP-Elites. An evolved map's quality is determined by its simplicity: the simpler it is, the better it is. Within this work, we show that the generated maps can strongly encourage or discourage different persona-like behaviors and range from simple solutions to complex puzzle-levels, making them perfect candidates for a tutorial generative system., Comment: 10 pages, 7 figures, 2 tables

We investigate how to efficiently predict play personas based on playtraces. Play personas can be computed by calculating the action agreement ratio between a player and a generative model of playing behavior, a so-called procedural persona. But this is computationally expensive and assumes that appropriate procedural personas are readily available. We present two methods for estimating player persona, one using regular supervised learning and aggregate measures of game mechanics initiated, and another based on sequence learning on a trace of closely cropped gameplay observations. While both of these methods achieve high accuracy when predicting play personas defined by agreement with procedural personas, they utterly fail to predict play style as defined by the players themselves using a questionnaire. This interesting result highlights the value of using computational methods in defining play personas., Comment: 8 pages, 3 tables, 2 figures

This article reviews some results of the SAGEX programme that have developed in the understanding of the interplay of supersymmetry and modular covariance of scattering amplitudes in type IIB superstring theory and its holographic image in $\mathcal{N}=4$ supersymmetric Yang-Mills theory (SYM). The first section includes the determination of exact expressions for BPS interactions in the low-energy expansion of type IIB superstring amplitudes. The second section concerns properties of a certain class of integrated correlators in $\mathcal{N}=4$ SYM with arbitrary classical gauge group that are exactly determined by supersymmetric localisation. Not only do these reproduce known features of perturbative and non-perturbative $\mathcal{N}=4$ SYM for any classical gauge group, but they have large-$N$ expansions that are in accord with expectations based on the holographic correspondence with superstring theory. The final section focusses on modular graph functions. These are modular functions that are closely associated with coefficients in the low-energy expansion of superstring perturbation theory and have recently received quite a lot of interest in both the physics and mathematics literature., Comment: 45 pages, see also the overview article arXiv:2203.13011. v3: journal version

This is an introduction to, and invitation to read, a series of review articles on scattering amplitudes in gauge theory, gravity, and superstring theory. Our aim is to provide an overview of the field, from basic aspects to a selection of current (2022) research and developments., Comment: 15 pages, overview article. v3: journal version

The purpose of this White Paper is to review recent progress towards elucidating and evaluating string amplitudes, relating them to quantum field theory amplitudes, applying their predictions to string dualities, exploring their connection with gravitational physics, and deepening our understanding of their mathematical structure. We also present a selection of targets for future research., Comment: 10 pages, contribution to Snowmass 2021, v2: references added

We present exact expressions for certain integrated correlators of four superconformal primary operators in the stress tensor multiplet of $\mathcal{N}=4$ supersymmetric Yang--Mills (SYM) theory with classical gauge group, $G_N$ $= SO(2N)$, $SO(2N+1)$, $USp(2N)$. These integrated correlators are expressed as two-dimensional lattice sums by considering derivatives of the localised partition functions, generalising the expression obtained for $SU(N)$ in our previous works. These expressions are manifestly covariant under Goddard-Nuyts-Olive duality. The integrated correlators can also be formally written as infinite sums of non-holomorphic Eisenstein series with integer indices and rational coefficients. Furthermore, the action of the hyperbolic Laplace operator with respect to the complex coupling $\tau=\theta/(2\pi) + 4\pi i /g^2_{_{YM}}$ on any integrated correlator for gauge group $G_N$ relates it to a linear combination of correlators with gauge groups $G_{N+1}$, $G_N$ and $G_{N-1}$. These "Laplace-difference equation" determine the expressions of integrated correlators for all classical gauge groups for any value of $N$ in terms of the correlator for the gauge group $SU(2)$. The perturbation expansions of these integrated correlators for any finite value of $N$ agree with properties obtained from perturbative Yang--Mills quantum field theory, together with various multi-instanton calculations which are also shown to agree with those determined by supersymmetric localisation. The coefficients of terms in the large-$N$ expansion are sums of non-holomorphic Eisenstein series with half-integer indices, which extend recent results and make contact with low order terms in the low energy expansion of type IIB superstring theory in an $AdS_5\times S^5/\mathbb{Z}_2$ background., Comment: 38 pages, v2: minor changes, version accepted for publication in SciPost

Rationales, snippets of extracted text that explain an inference, have emerged as a popular framework for interpretable natural language processing (NLP). Rationale models typically consist of two cooperating modules: a selector and a classifier with the goal of maximizing the mutual information (MMI) between the "selected" text and the document label. Despite their promises, MMI-based methods often pick up on spurious text patterns and result in models with nonsensical behaviors. In this work, we investigate whether counterfactual data augmentation (CDA), without human assistance, can improve the performance of the selector by lowering the mutual information between spurious signals and the document label. Our counterfactuals are produced in an unsupervised fashion using class-dependent generative models. From an information theoretic lens, we derive properties of the unaugmented dataset for which our CDA approach would succeed. The effectiveness of CDA is empirically evaluated by comparing against several baselines including an improved MMI-based rationale schema on two multi aspect datasets. Our results show that CDA produces rationales that better capture the signal of interest.

The exact expressions for integrated maximal $U(1)_Y$ violating (MUV) $n$-point correlators in $SU(N)$ ${\mathcal N}=4$ supersymmetric Yang--Mills theory are determined. The analysis generalises previous results on the integrated correlator of four superconformal primaries and is based on supersymmetric localisation. The integrated correlators are functions of $N$ and $\tau=\theta/(2\pi)+4\pi i/g_{_{YM}}^2$, and are expressed as two-dimensional lattice sums that are modular forms with holomorphic and anti-holomorphic weights $(w,-w)$ where $w=n-4$. The correlators satisfy Laplace-difference equations that relate the $SU(N+1)$, $SU(N)$ and $SU(N-1)$ expressions and generalise the equations previously found in the $w=0$ case. The correlators can be expressed as infinite sums of Eisenstein modular forms of weight $(w,-w)$. For any fixed value of $N$ the perturbation expansion of this correlator is found to start at order $( g_{_{YM}}^2 N)^w$. The contributions of Yang--Mills instantons of charge $k>0$ are of the form $q^k\, f(g_{_{YM}})$, where $q=e^{2\pi i \tau}$ and $f(g_{_{YM}})= O(g_{_{YM}}^{-2w})$ when $g_{_{YM}}^2 \ll 1$. Anti-instanton contributions have charge $k<0$ and are of the form $\bar q^{|k|} \, \hat f(g_{_{YM}})$, where $\hat f(g_{_{YM}}) = O(g_{_{YM}}^{2w})$ when $g_{_{YM}}^2 \ll 1$. Properties of the large-$N$ expansion are in agreement with expectations based on the low energy expansion of flat-space type IIB superstring amplitudes. We also comment on the identification of $n$-point free-field MUV correlators with the integrands of $(n-4)$-loop perturbative contributions to the four-point correlator. In particular, we emphasise the important r\^ole of $SL(2, \mathbb{Z})$-covariance in the construction., Comment: 40 pages. v2 minor typos corrected

The first experimental demonstration of an oscillator based on a periodic, resonant microstrip circuit with a degenerate band edge (DBE) is presented. The DBE is a fourth-order exceptional degeneracy of the eigenmodes in a lossless periodic waveguide that is visible in the wavenumber-frequency dispersion diagram, and the periodic microstrip behaves as a frequency selective medium. The presence of the DBE condition and the associated DBE resonance allow for a stable, single-mode oscillation as well as stability with varying the load resistance.

The GDMC AI settlement generation challenge is a PCG competition about producing an algorithm that can create an "interesting" Minecraft settlement for a given map. This paper contains a collection of written experiences with this competition, by participants, judges, organizers and advisors. We asked people to reflect both on the artifacts themselves, and on the competition in general. The aim of this paper is to offer a shareable and edited collection of experiences and qualitative feedback - which seem to contain a lot of insights on PCG and computational creativity, but would otherwise be lost once the output of the competition is reduced to scalar performance values. We reflect upon some organizational issues for AI competitions, and discuss the future of the GDMC competition., Comment: 28 pages, 5 figures

We derive the equations for the odd and even parity perturbations of coupled electromagnetic and gravitational fields of a black hole with an electric charge within the context of general nonlinear electrodynamics. The Lagrangian density is a generic function of the Lorentz invariant scalar quantities of the electromagnetic fields. We include the Hodge dual of the electromagnetic field tensor and the cosmological constant in our calculations. For each type of parity, we reduce the system of Einstein field equations coupled to nonlinear electrodynamics to two coupled Schr\"odinger-type wave equations, one for the gravitational field and one for the electromagnetic field. The stability conditions in the presence of the Hodge dual of the electromagnetic field are derived., Comment: 25 pages. Minor corrections are made and one figure is added. The paper is going to be published in Phys. Rev. D soon