Search

Model merging has gained prominence in machine learning as a method to integrate multiple trained models into a single model without accessing the original training data. While existing approaches have demonstrated success in domains such as computer vision and NLP, their application to Graph Neural Networks (GNNs) remains unexplored. These methods often rely on the assumption of shared initialization, which is seldom applicable to GNNs. In this work, we undertake the first benchmarking study of model merging algorithms for GNNs, revealing their limited effectiveness in this context. To address these challenges, we propose GNNMerge, which utilizes a task-agnostic node embedding alignment strategy to merge GNNs. Furthermore, we establish that under a mild relaxation, the proposed optimization objective admits direct analytical solutions for widely used GNN architectures, significantly enhancing its computational efficiency. Empirical evaluations across diverse datasets, tasks, and architectures establish GNNMerge to be up to 24% more accurate than existing methods while delivering over 2 orders of magnitude speed-up compared to training from scratch.

As short text data in native languages like Hindi increasingly appear in modern media, robust methods for topic modeling on such data have gained importance. This study investigates the performance of BERTopic in modeling Hindi short texts, an area that has been under-explored in existing research. Using contextual embeddings, BERTopic can capture semantic relationships in data, making it potentially more effective than traditional models, especially for short and diverse texts. We evaluate BERTopic using 6 different document embedding models and compare its performance against 8 established topic modeling techniques, such as Latent Dirichlet Allocation (LDA), Non-negative Matrix Factorization (NMF), Latent Semantic Indexing (LSI), Additive Regularization of Topic Models (ARTM), Probabilistic Latent Semantic Analysis (PLSA), Embedded Topic Model (ETM), Combined Topic Model (CTM), and Top2Vec. The models are assessed using coherence scores across a range of topic counts. Our results reveal that BERTopic consistently outperforms other models in capturing coherent topics from short Hindi texts., Comment: Accepted into IndoNLP: The First Workshop on Natural Language Processing for Indo-Aryan and Dravidian Languages, collocated with COLING 2025. Set to appear in the workshop proceedings published in ACL Anthology

This paper offers a comprehensive analysis of recent advancements in video inpainting techniques, a critical subset of computer vision and artificial intelligence. As a process that restores or fills in missing or corrupted portions of video sequences with plausible content, video inpainting has evolved significantly with the advent of deep learning methodologies. Despite the plethora of existing methods and their swift development, the landscape remains complex, posing challenges to both novices and established researchers. Our study deconstructs major techniques, their underpinning theories, and their effective applications. Moreover, we conduct an exhaustive comparative study, centering on two often-overlooked dimensions: visual quality and computational efficiency. We adopt a human-centric approach to assess visual quality, enlisting a panel of annotators to evaluate the output of different video inpainting techniques. This provides a nuanced qualitative understanding that complements traditional quantitative metrics. Concurrently, we delve into the computational aspects, comparing inference times and memory demands across a standardized hardware setup. This analysis underscores the balance between quality and efficiency: a critical consideration for practical applications where resources may be constrained. By integrating human validation and computational resource comparison, this survey not only clarifies the present landscape of video inpainting techniques but also charts a course for future explorations in this vibrant and evolving field.

The COVID19 pandemic created a worldwide emergency as it is estimated that such a large number of infections are due to human-to-human transmission of the COVID19. As a necessity, there is a need to track users who came in contact with users having travel history, asymptomatic and not yet symptomatic, but they can be in the future. To solve this problem, the present work proposes a solution for contact tracing based on assisted GPS and cloud computing technologies. An application is developed to collect each user's assisted GPS coordinates once all the users install this application. This application periodically sends assisted GPS data to the cloud. To determine which devices are within the permissible limit of 5m, we perform clustering over assisted GPS coordinates and track the clusters for about t mins to allow the measure of spread. We assume that it takes around 3 or 5 mins to get the virus from an infected object. For clustering, the proposed M way like tree data structure stores the assisted GPS coordinates in degree, minute, and second format. Thus, every user is mapped to a leaf node of the tree. We split the "seconds" part of the assisted GPS location into m equal parts, which amount to d meter in latitude(longitude). Hence, two users who are within d meter range will map to the same leaf node. Thus, by mapping assisted GPS locations every t mins, we can find out how many users came in contact with a particular user for at least t mins. Our work's salient feature is that it runs in linear time O(n) for n users in the static case, i.e., when users are not moving. We also propose a variant of our solution to handle the dynamic case, that is, when users are moving. Besides, the proposed solution offers potential hotspot detection and safe-route recommendation as an additional feature, and proof of concept is presented through experiments on simulated data of 10M users., Comment: 10 pages

This paper establishes a close relationship among the four information theoretic problems, namely Campbell source coding, Arikan guessing, Huleihel et al. memoryless guessing and Bunte and Lapidoth tasks partitioning problems. We first show that the aforementioned problems are mathematically related via a general moment minimization problem whose optimum solution is given in terms of Renyi entropy. We then propose a general framework for the mismatched version of these problems and establish all the asymptotic results using this framework. Further, we study an ordered tasks partitioning problem that turns out to be a generalisation of Arikan's guessing problem. Finally, with the help of this general framework, we establish an equivalence among all these problems, in the sense that, knowing an asymptotically optimal solution in one problem helps us find the same in all other problems., Comment: 24 pages, 1 figure. arXiv admin note: text overlap with arXiv:1907.06889

We introduce Mitiq, a Python package for error mitigation on noisy quantum computers. Error mitigation techniques can reduce the impact of noise on near-term quantum computers with minimal overhead in quantum resources by relying on a mixture of quantum sampling and classical post-processing techniques. Mitiq is an extensible toolkit of different error mitigation methods, including zero-noise extrapolation, probabilistic error cancellation, and Clifford data regression. The library is designed to be compatible with generic backends and interfaces with different quantum software frameworks. We describe Mitiq using code snippets to demonstrate usage and discuss features and contribution guidelines. We present several examples demonstrating error mitigation on IBM and Rigetti superconducting quantum processors as well as on noisy simulators., Comment: 33 pages, 8 figures. The Mitiq GitHub is https://github.com/unitaryfund/mitiq and the Mitiq documentation is https://mitiq.readthedocs.io/en/stable/

We study four problems namely, Campbell's source coding problem, Arikan's guessing problem, Huieihel et al.'s memoryless guessing problem, and Bunte and Lapidoth's task partitioning problem. We observe a close relationship among these problems. In all these problems, the objective is to minimize moments of some functions of random variables, and R\'enyi entropy and Sundaresan's divergence arise as optimal solutions. This motivates us to establish a connection among these four problems. In this paper, we study a more general problem and show that R\'{e}nyi and Shannon entropies arise as its solution. We show that the problems on source coding, guessing and task partitioning are particular instances of this general optimization problem, and derive the lower bounds using this framework. We also refine some known results and present new results for mismatched version of these problems using a unified approach. We strongly feel that this generalization would, in addition to help in understanding the similarities and distinctiveness of these problems, also help to solve any new problem that falls in this framework., Comment: 22 pages

We review all the published literature and show that there is no experimental evidence for homogeneous tin titanate SnTiO3 in bulk or thin-film form. Instead a combination of unrelated artefacts are easily misinterpreted. The X-ray Bragg data are contaminated by double scattering from the Si substrate, giving a strong line at the 2-theta angle exactly where perovskite SnTiO3 should appear. The strong dielectric divergence near 560K is irreversible and arises from oxygen site detrapping, accompanied by Warburg/Randles interfacial anomalies. The small (4 uC/cm2) apparent ferroelectric hysteresis remains in samples shown in pure (Sn,Ti)O2 rutile/cassiterite, in which ferroelectricity is forbidden. Only very recent German work reveals real bulk SnTiO3, but this is completely inhomogeneous, consisting of an elaborate array of stacking faults, not suitable for ferroelectric devices. Unpublished TEM data reveal an inhomogeneous SnO layered structured thin films, related to shell-core structures. The harsh conclusion is that there is a combination of unrelated artefacts masquerading as ferroelectricity in powders and ALD films; and only a trace of a second phase in Cambridge PLD data suggests any perovskite content at all. The fact that X-ray, dielectric, and hysteresis data all lead to the wrong conclusion is instructive and reminds us of earlier work on copper calcium titanate (a well-known boundary-layer capacitor)., Comment: "This is the version of the article before peer review or editing, as submitted by an author to Reports on Progress in Physics. IOP Publishing Ltd is not responsible for any errors or omissions in this version of the manuscript or any version derived from it. The Version of Record is available online at https://dx.doi.org/10.1088/1361-6633/ab37d4."

We investigate an obstruction for hypersymplectic manifolds equipped with a free, isometric action of SU(1,1). When the obstruction vanishes, we show that the manifold is a metric cone over a split 3-Sasakian manifold. Furthermore, if the action of SU(1,1) is also proper, then the hypersymplectic manifold fibres over a para-quaternionic Kahler manifold. We conclude the article with some examples for which the obstruction vanishes. In particular, we show that the moduli space to Nahm-Schmid equations admits a fibration over a para-quaternionic Kahler manifold., Comment: 21 pages. Some more minor corrections. (To appear in Q. J. Math.)

We obtain a Hitchin-Kobayashi-type correspondence for symplectic vortex equations, with the target a Kahler cone over a compact Sasakian manifold. We show that the correspondence reduces to studying the existence and uniqueness of Kazdan-Warner equations. Using this, we construct a map between the moduli space of solutions to the symplectic vortex equations and effective divisors., Comment: 9 pages. Comments/ suggestions welcome!

In this article, we establish a Hitchin-Kobayashi type correspondence for generalised Seiberg-Witten monopole equations on Kahler surfaces. We show that the "stability" criterion we obtain, for the existence of solutions, coincides with that of the usual Seiberg-Witten monopole equations. This enables us to construct a map from the moduli space of solutions to the generalised equations to effective divisors., Comment: 14 pages, minor corrections. To appear in Journal of Mathematical Physics

In this article, we study a generalisation of the Seiberg-Witten equations, replacing the spinor representation with a hyperKahler manifold equipped with certain symmetries. Central to this is the construction of a (non-linear) Dirac operator acting on the sections of the non-linear fibre-bundle. For hyperKahler manifolds admitting a hyperKahler potential, we derive a transformation formula for the Dirac operator under the conformal change of metric on the base manifold. As an application, we show that when the hyperKahler manifold is of dimension four, then away from a singular set, the equations can be expressed as a second order PDE in terms of almost-complex structure on the base manifold and a conformal factor. This extends a result of Donaldson to generalised Seiberg-Witten equations., Comment: 21 pages. Several corrections, especially in the second half of the article, and much improved exposition. Accepted at J. Geom. Phys

We report giant resistive switching of an order of 104, long-time charge retention characteristics up to 104 s, non-overlapping SET and RESET voltages, ohmic in low resistance state (LRS) and space charge limited current (SCLC) mechanism in high resistance state (HRS) properties in polycrystalline perovskite Cobalt Titanate (CoTiO3 ~ CTO) thin films. Impedance spectroscopy study was carried out for both LRS and HRS states which illustrates that only bulk resistance changes after resistance switching, however, there is a small change (<10% which is in pF range) in the bulk capacitance value in both states. These results suggest that in LRS state current filaments break the capacitor in many small capacitors in a parallel configuration which in turn provides the same capacitance in both states even there was 90 degree changes in phase-angle and an order of change in the tangent loss., Comment: 7 pages, 4 figures

A non-linear generalization of the Dirac operator in 4-dimensions, obtained by replacing the spinor representation with a hyperKahler manifold admitting certain symmetries, is considered. We show that the existence of a covariantly constant, generalized spinor defines a Kahler structure on the base 4-dimensional manifold. For a class of hyperKahler manifolds obtained via hyperKahler reduction, we also show that a harmonic spinor, under mild conditions, defines a symplectic structure. Finally, we show that if a covariantly constant, generalized spinor satisfies generalized Seiberg-Witten equations, the metric on the base manifold has a constant scalar curvature., Comment: 15 pages, added the case of hyperKahler manifolds obtained via hyperKahler reduction and rectified a couple of embarrassing mistakes in the previous version of the paper. Comments and suggestions are welcome!