Search

The deep operator network (DeepONet) is a popular neural operator architecture that has shown promise in solving partial differential equations (PDEs) by using deep neural networks to map between infinite-dimensional function spaces. In the absence of labeled datasets, we utilize the PDE residual loss to learn the physical system, an approach known as physics-informed DeepONet. This method faces significant computational challenges, primarily due to the curse of dimensionality, as the computational cost increases exponentially with finer discretization. In this paper, we introduce the Separable DeepONet framework to address these challenges and improve scalability for high-dimensional PDEs. Our approach involves a factorization technique where sub-networks handle individual one-dimensional coordinates, thereby reducing the number of forward passes and the size of the Jacobian matrix. By using forward-mode automatic differentiation, we further optimize the computational cost related to the Jacobian matrix. As a result, our modifications lead to a linear scaling of computational cost with discretization density, making Separable DeepONet suitable for high-dimensional PDEs. We validate the effectiveness of the separable architecture through three benchmark PDE models: the viscous Burgers equation, Biot's consolidation theory, and a parametrized heat equation. In all cases, our proposed framework achieves comparable or improved accuracy while significantly reducing computational time compared to conventional DeepONet. These results demonstrate the potential of Separable DeepONet in efficiently solving complex, high-dimensional PDEs, advancing the field of physics-informed machine learning., Comment: 23 Pages, 9 Figures and 1 Table

Distributed quantum computing leverages the collective power of multiple quantum devices to perform computations exceeding the capabilities of individual quantum devices. A currently studied technique to enable this distributed approach is wire cutting, which decomposes a quantum circuit into smaller subcircuits by cutting their connecting wires. These subcircuits can then be executed on distributed devices, and their results are classically combined to reconstruct the original computation's result. However, wire cutting requires additional circuit executions to preserve result accuracy, with their number growing exponentially with each cut. Thus, minimizing this sampling overhead is crucial for reducing the total execution time. Employing shared non-maximally entangled (NME) states between distributed devices reduces this overhead for single wire cuts, moving closer to ideal teleportation with maximally entangled states. Extending this approach to jointly cutting multiple wires using NME states remained unexplored. Our paper addresses this gap by investigating the use of NME states for joint wire cuts, aiming to reduce the sampling overhead further. Our three main contributions include (i) determining the minimal sampling overhead for this scenario, (ii) analyzing the overhead when using composite NME states constructed from smaller NME states, and (iii) introducing a wire cutting technique that achieves the optimal sampling overhead with pure NME states, paving the way towards wire cutting with arbitrary NME states., Comment: 33 pages, 6 figures

Cover songs are alternate versions of a song by a different artist. Long being a vital part of the music industry, cover songs significantly influence music culture and are commonly heard in public venues. The rise of online music platforms has further increased their prevalence, often as background music or video soundtracks. While current automatic identification methods serve adequately for original songs, they are less effective with cover songs, primarily because cover versions often significantly deviate from the original compositions. In this paper, we propose a novel method for cover song detection that utilizes the lyrics of a song. We introduce a new dataset for cover songs and their corresponding originals. The dataset contains 5078 cover songs and 2828 original songs. In contrast to other cover song datasets, it contains the annotated lyrics for the original song and the cover song. We evaluate our method on this dataset and compare it with multiple baseline approaches. Our results show that our method outperforms the baseline approaches., Comment: 6 pages, 3 figures

Distributed quantum computing supports combining the computational power of multiple quantum devices to overcome the limitations of individual devices. Circuit cutting techniques enable the distribution of quantum computations via classical communication. These techniques involve partitioning a quantum circuit into smaller subcircuits, each containing fewer qubits. The original circuit's outcome can be replicated by executing these subcircuits on separate devices and combining their results. However, the number of circuit executions required to achieve a fixed result accuracy with circuit cutting grows exponentially with the number of cuts, posing significant costs. In contrast, quantum teleportation allows the distribution of quantum computations without an exponential increase in circuit executions. Nevertheless, each teleportation requires a pre-shared pair of maximally entangled qubits for transmitting a quantum state, and non-maximally entangled qubits cannot be used for this purpose. Addressing this, our work explores utilizing non-maximally entangled qubit pairs in wire cutting, a specific form of circuit cutting, to mitigate the associated costs. The cost of this cutting procedure reduces with the increasing degree of entanglement in the pre-shared qubit pairs. We derive the optimal sampling overhead in this context and present a wire cutting technique employing pure non-maximally entangled states that achieves this optimal sampling overhead. Hence, this offers a continuum between existing wire cutting and quantum teleportation., Comment: 10 pages, 7 figures. This article supersedes arXiv:2306.12084. Accepted in the 2024 IPDPS Workshop on Quantum Computing Algorithms, Systems, and Applications

We contribute the first publicly available dataset of factual claims from different platforms and fake YouTube videos on the 2023 Israel-Hamas war for automatic fake YouTube video classification. The FakeClaim data is collected from 60 fact-checking organizations in 30 languages and enriched with metadata from the fact-checking organizations curated by trained journalists specialized in fact-checking. Further, we classify fake videos within the subset of YouTube videos using textual information and user comments. We used a pre-trained model to classify each video with different feature combinations. Our best-performing fine-tuned language model, Universal Sentence Encoder (USE), achieves a Macro F1 of 87\%, which shows that the trained model can be helpful for debunking fake videos using the comments from the user discussion. The dataset is available on Github\footnote{https://github.com/Gautamshahi/FakeClaim}, Comment: Accepted in the IR4Good Track at the 46th European Conference on Information Retrieval (ECIR) 2024

When different researchers study the same research question using the same dataset they may obtain different and potentially even conflicting results. This is because there is often substantial flexibility in researchers' analytical choices, an issue also referred to as ''researcher degrees of freedom''. Combined with selective reporting of the smallest p-value or largest effect, researcher degrees of freedom may lead to an increased rate of false positive and overoptimistic results. In this paper, we address this issue by formalizing the multiplicity of analysis strategies as a multiple testing problem. As the test statistics of different analysis strategies are usually highly dependent, a naive approach such as the Bonferroni correction is inappropriate because it leads to an unacceptable loss of power. Instead, we propose using the ''minP'' adjustment method, which takes potential test dependencies into account and approximates the underlying null distribution of the minimal p-value through a permutation-based procedure. This procedure is known to achieve more power than simpler approaches while ensuring a weak control of the family-wise error rate. We illustrate our approach for addressing researcher degrees of freedom by applying it to a study on the impact of perioperative paO2 on post-operative complications after neurosurgery. A total of 48 analysis strategies are considered and adjusted using the minP procedure. This approach allows to selectively report the result of the analysis strategy yielding the most convincing evidence, while controlling the type 1 error -- and thus the risk of publishing false positive results that may not be replicable.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

We study the (2+1)-dimensional Gross-Neveu model at non-zero chemical potential and subjected to a homogeneous background magnetic field. We do so both analytically, in the limit of an infinite number of fermion flavors in which mean-field approaches become exact, as well as on the lattice for a single flavor. The rich and exotic phase structure observed in the mean-field limit is found to be destroyed when strong quantum fluctuations are present in the system. Instead, in the phase of spontaneously broken chiral symmetry the magnetic field enhances this breaking for all choices of parameters. As a byproduct, we find indications for a first-order phase transition in the chemical potential for vanishing magnetic field but also provide hints that this could rather be a finite-size than a finite-flavor-number effect., Comment: 6 pages, 4 figures; Proceedings of the 40th International Symposium on Lattice Field Theory (Lattice 2023)

Training classical neural networks generally requires a large number of training samples. Using entangled training samples, Quantum Neural Networks (QNNs) have the potential to significantly reduce the amount of training samples required in the training process. However, to minimize the number of incorrect predictions made by the resulting QNN, it is essential that the structure of the training samples meets certain requirements. On the one hand, the exact degree of entanglement must be fixed for the whole set of training samples. On the other hand, training samples must be linearly independent and non-orthogonal. However, how failing to meet these requirements affects the resulting QNN is not fully studied. To address this, we extend the proof of the QNFL theorem to (i) provide a generalization of the theorem for varying degrees of entanglement. This generalization shows that the average degree of entanglement in the set of training samples can be used to predict the expected quality of the QNN. Furthermore, we (ii) introduce new estimates for the expected accuracy of QNNs for moderately entangled training samples that are linear dependent or orthogonal. Our analytical results are (iii) experimentally validated by simulating QNN training and analyzing the quality of the QNN after training., Comment: 35 pages, 6 figures; changed layout and fixed typos

Distributed quantum computing combines the computational power of multiple devices to overcome the limitations of individual devices. Circuit cutting techniques enable the distribution of quantum computations through classical communication. These techniques involve partitioning a quantum circuit into smaller subcircuits, each containing fewer qubits. The original circuit's outcome can be replicated by executing these subcircuits on separate devices and combining their results. However, the number of shots required to achieve a fixed result accuracy with circuit cutting grows exponentially with the number of cuts, posing significant costs. In contrast, quantum teleportation allows the distribution of quantum computations without an exponential increase in shots. Nevertheless, each teleportation procedure requires a pre-shared pair of maximally entangled qubits for transmitting a quantum state, and non-maximally entangled qubits cannot be used for this purpose. To address this, we propose a novel circuit cutting technique that leverages non-maximally entangled qubit pairs, effectively reducing the cost associated with circuit cutting. By considering the degree of entanglement in the pre-shared qubit pairs, our technique provides a continuum between existing circuit cutting methods and quantum teleportation, adjusting the cost of circuit cutting accordingly., Comment: 9 pages, 4 figures

We perform a lattice study of the ($2+1$)-dimensional Gross-Neveu model in a background magnetic field $B$ and at non-zero chemical potential $\mu$. The complex-action problem arising in our simulations using overlap fermions is under control. For $B=0$ we observe a first-order phase transition in $\mu$ even at non-vanishing temperatures. Our main finding, however, is that the rich phase structure found in the limit of infinite flavor number $N_\mathrm{f}$ is washed out by the fluctuations present at $N_\mathrm{f}=1$. We find no evidence for inverse magnetic catalysis, i.e., the decrease of the order parameter of chiral symmetry breaking with $B$ for $\mu$ close to the chiral phase transition. Instead, the magnetic field tends to enhance the breakdown of chiral symmetry for all values of $\mu$ below the transition. Moreover, we find no trace of spatial inhomogeneities in the order parameter. We briefly comment on the potential relevance of our results for QCD., Comment: 8 pages + 1 page appendix, 7 figures, version published in PRD

Due to low numbers of qubits and their error-proneness, Noisy Intermediate-Scale Quantum (NISQ) computers impose constraints on the size of quantum algorithms they can successfully execute. State-of-the-art research introduces various techniques addressing these limitations by utilizing known or inexpensively generated approximations, solutions, or models as a starting point to approach a task instead of starting from scratch. These so-called warm-starting techniques aim to reduce quantum resource consumption, thus facilitating the design of algorithms suiting the capabilities of NISQ computers. In this work, we collect and analyze scientific literature on warm-starting techniques in the quantum computing domain. In particular, we (i) create a systematic map of state-of-the-art research on warm-starting techniques using established guidelines for systematic mapping studies, (ii) identify relevant properties of such techniques, and (iii) based on these properties classify the techniques identified in the literature in an extensible classification scheme. Our results provide insights into the research field and aim to help quantum software engineers to categorize warm-starting techniques and apply them in practice. Moreover, our contributions may serve as a starting point for further research on the warm-starting topic since they provide an overview of existing work and facilitate the identification of research gaps., Comment: 29 pages, 8 figures, 5 tables, This is the authors' version of the work. It is posted here for your personal use. Not for redistribution. The definitive version was published in ACM Computing Surveys, https://doi.org/10.1145/3652510

The Boolean satisfiability problem (SAT), in particular 3SAT with its bounded clause size, is a well-studied problem since a wide range of decision problems can be reduced to it. Due to its high complexity, examining potentials of quantum algorithms for solving 3SAT more efficiently is an important topic. Since 3SAT can be formulated as unstructured search for satisfying variable assignments, the amplitude amplification algorithm can be applied. However, the high circuit complexity of amplitude amplification hinders its use on near-term quantum systems. On the other hand, the Quantum Approximate Optimization Algorithm (QAOA) is a promising candidate for solving 3SAT for Noisy Intermediate-Scale Quantum devices in the near future due to its simple quantum ansatz. However, although QAOA generally exhibits a high approximation ratio, there are 3SAT problem instances where its success probability decreases using current implementations. To address this problem, in this paper we introduce amplitude amplification-inspired variants of QAOA to improve the success probability for 3SAT. For this, (i) three amplitude amplification-inspired QAOA variants are introduced and implemented, (ii) the variants are experimental compared with a standard QAOA implementation, and (iii) the impact on the success probability and ansatz complexity is analyzed. The experiment results show that an improvement in the success probability can be achieved with only a moderate increase in circuit complexity.

We study the Gross-Neveu model in $2+1$ dimensions in an external magnetic field $B$. We first summarize known mean-field results, obtained in the limit of large flavor number $N_\mathrm{f}$, before presenting lattice results using the overlap discretization to study one reducible fermion flavor, $N_\mathrm{f}=1$. Our findings indicate that the magnetic catalysis phenomenon, i.e., an increase of the chiral condensate with the magnetic field, persists beyond the mean-field limit for temperatures below the chiral phase transition and that the critical temperature grows with increasing magnetic field. This is in contrast to the situation in QCD, where the broken phase shrinks with increasing $B$ while the condensate exhibits a non-monotonic $B$-dependence close to the chiral crossover, and we comment on this discrepancy. We do not find any trace of inhomogeneous phases induced by the magnetic field., Comment: 10 pages + 4 pages appendix, 10 figures; some clarifications, corrected typos, added references

Noisy Intermediate-Scale Quantum (NISQ) devices are restricted by their limited number of qubits and their short decoherence times. An approach addressing these problems is quantum circuit cutting. It decomposes the execution of a large quantum circuit into the execution of multiple smaller quantum circuits with additional classical postprocessing. Since these smaller quantum circuits require fewer qubits and gates, they are more suitable for NISQ devices. To investigate the effect of quantum circuit cutting in a quantum algorithm targeting NISQ devices, we design two experiments using the Quantum Approximate Optimization Algorithm (QAOA) for the Maximum Cut (MaxCut) problem and conduct them on state-of-the-art superconducting devices. Our first experiment studies the influence of circuit cutting on the objective function of QAOA, and the second evaluates the quality of results obtained by the whole algorithm with circuit cutting. The results show that circuit cutting can reduce the effects of noise in QAOA, and therefore, the algorithm yields better solutions on NISQ devices.

We investigate the Gross-Neveu model in $2+1$ dimensions in a constant and homogeneous magnetic field using one reducible flavor of overlap fermions. Our lattice simulations suggest that the magnetic catalysis phenomenon, i.e., an increase of the chiral condensate with the magnetic field, is present for all temperatures below the chiral phase transition, in accordance with predictions from mean-field and beyond-mean-field calculations., Comment: 7 pages, 4 figures; talk given at the 39th International Symposium on Lattice Field theory (LATTICE 2022) in Bonn; August 8-13 2022

The widespread of offensive content online such as hate speech poses a growing societal problem. AI tools are necessary for supporting the moderation process at online platforms. For the evaluation of these identification tools, continuous experimentation with data sets in different languages are necessary. The HASOC track (Hate Speech and Offensive Content Identification) is dedicated to develop benchmark data for this purpose. This paper presents the HASOC subtrack for English, Hindi, and Marathi. The data set was assembled from Twitter. This subtrack has two sub-tasks. Task A is a binary classification problem (Hate and Not Offensive) offered for all three languages. Task B is a fine-grained classification problem for three classes (HATE) Hate speech, OFFENSIVE and PROFANITY offered for English and Hindi. Overall, 652 runs were submitted by 65 teams. The performance of the best classification algorithms for task A are F1 measures 0.91, 0.78 and 0.83 for Marathi, Hindi and English, respectively. This overview presents the tasks and the data development as well as the detailed results. The systems submitted to the competition applied a variety of technologies. The best performing algorithms were mainly variants of transformer architectures.

We study the $(1+1)$-dimensional chiral Gross-Neveu model on the lattice. At finite density, analytic mean-field results predict the existence of inhomogeneous condensates breaking both chiral symmetry and spacetime symmetries spontaneously. We investigate the fate of these inhomogeneities for two flavors and find remnant structural order, albeit with a decaying amplitude. We also map out phase diagrams in the plane spanned by the chemical potential and temperature for different lattice spacings and physical volumes. Finally, we comment on the interpretation of our results in the light of various no-go theorems., Comment: 5 pages, 5 figures

We describe the fourth edition of the CheckThat! Lab, part of the 2021 Conference and Labs of the Evaluation Forum (CLEF). The lab evaluates technology supporting tasks related to factuality, and covers Arabic, Bulgarian, English, Spanish, and Turkish. Task 1 asks to predict which posts in a Twitter stream are worth fact-checking, focusing on COVID-19 and politics (in all five languages). Task 2 asks to determine whether a claim in a tweet can be verified using a set of previously fact-checked claims (in Arabic and English). Task 3 asks to predict the veracity of a news article and its topical domain (in English). The evaluation is based on mean average precision or precision at rank k for the ranking tasks, and macro-F1 for the classification tasks. This was the most popular CLEF-2021 lab in terms of team registrations: 132 teams. Nearly one-third of them participated: 15, 5, and 25 teams submitted official runs for tasks 1, 2, and 3, respectively., Comment: Check-Worthiness Estimation, Fact-Checking, Veracity, Evidence-based Verification, Detecting Previously Fact-Checked Claims, Social Media Verification, Computational Journalism, COVID-19

We investigate the finite-temperature and -density chiral Gross-Neveu model with an axial U$_A$(1) symmetry in $1+1$ dimensions on the lattice. In the limit where the number of flavors $N_\mathrm{f}$ tends to infinity the continuum model has been solved analytically and shows two phases: a symmetric high-temperature phase with a vanishing condensate and a low-temperature phase in which the complex condensate forms a chiral spiral which breaks translation invariance. In the lattice simulations we employ chiral SLAC fermions with exact axial symmetry. Similarly to $N_\mathrm{f}\to\infty$, we find for $8$ flavors, where quantum and thermal fluctuations are suppressed, two distinct regimes in the $(T,\mu)$ phase diagram, characterized by qualitatively different behavior of the two-point functions of the condensate fields. More surprisingly, at $N_\mathrm{f}=2$, where fluctuations are no longer suppressed, the model still behaves similarly to the $N_\mathrm{f}\to\infty$ model and we conclude that the chiral spiral leaves its footprints even on systems with a small number of flavors. For example, at low temperature the two-point functions are still dominated by chiral spirals with pitches proportional to the inverse chemical potential, although in contrast to large-$N_\mathrm{f}$ their amplitudes decrease with distance. We argue that these results should not be interpreted as the spontaneous breaking of a continuous symmetry, which is forbidden in two dimensions. Finally, using Dyson-Schwinger equations we calculate the decay of the U$_A$(1)-invariant fermion four-point function in search for a BKT phase at zero temperature., Comment: 15 pages + 5 pages appendix, 16 figures; added appendix on autocorrelations, added ancillary files, added clarifications