Search

In the past few years, the field of quantum computing is reaching new heights with significant advancements in algorithm development. In parallel to rising research areas, companies and research labs are actively working to build fault-tolerant quantum computers which can help provide accurate and speedy results for the various experiments. The increasing demand for quantum computers necessitates the sharing of hardware to enable multi-tenancy for a broad user base. While this approach optimizes the utilization of limited quantum resources, it also introduces potential security vulnerabilities. In this paper we examine dynamical decoupling (DD) as a countermeasure to protect the legitimate circuit from such threats. We focus on crosstalk-mediated attacks on Grover's search algorithm. We find that, when compared to other countermeasures, DD successfully mitigates the attack and in some cases is able to improve the performance of the circuit beyond the level of no-attack. Thus our results emphasis the importance of incorporating DD into algorithm executions on multi-tenancy quantum hardware.

The $1/c$-expansion of general relativity appropriately sourced by matter can be used to derive an action principle for Newtonian gravity. The gravitational part of this action is known as non-relativistic gravity (NRG). It is possible to source NRG differently and in such a way that one can construct solutions that are not described by Newtonian gravity (as they do not admit a notion of absolute time). It is possible to include a negative cosmological constant such that NRG admits a non-relativistic AdS solution. This non-relativistic AdS vacuum has Killing vectors that form the Galilean conformal algebra and a boundary that admits a conformal class of Newton-Cartan geometries. This begs the question of whether there exists an analogue of the fluid/gravity correspondence for NRG. In this paper we derive a non-relativistic AdS brane solution of NRG and confirm that it corresponds to the $1/c^2$-expansion of the AdS black brane geometry. We perform a Galilean boost of the non-relativistic AdS brane and derive the associated boundary energy-momentum tensor. We then show that this is the energy-momentum tensor of a massless Galilean fluid and explain how this is linked to the conformal isometries of the boundary. Along the way, we also present several new results for the theory of non-relativistic gravity itself. In particular we present a rewriting that greatly shortens and simplifies the equations of motion of the NRG action., Comment: 45 pages, 4 appendices

The focus of this paper is to increase our understanding of the Longest Processing Time First (LPT) heuristic. LPT is a classical heuristic for the fundamental problem of uniform machine scheduling. For different machine speeds, LPT was first considered by Gonzalez et al (SIAM J. Computing, 1977). Since then, extensive work has been done to improve the approximation factor of the LPT heuristic. However, all known implementations of the LPT heuristic take $O(mn)$ time, where $m$ is the number of machines and $n$ is the number of jobs. In this work, we come up with the first near-linear time implementation for LPT. Specifically, the running time is $O((n+m)(\log^2{m}+\log{n}))$. Somewhat surprisingly, the result is obtained by mapping the problem to dynamic maintenance of lower envelope of lines, which has been well studied in the computational geometry community. Our second contribution is to analyze the performance of LPT for the Drones Warehouse Problem (DWP), which is a natural generalization of the uniform machine scheduling problem motivated by drone-based parcel delivery from a warehouse. In this problem, a warehouse has multiple drones and wants to deliver parcels to several customers. Each drone picks a parcel from the warehouse, delivers it, and returns to the warehouse (where it can also get charged). The speeds and battery lives of the drones could be different, and due to the limited battery life, each drone has a bounded range in which it can deliver parcels. The goal is to assign parcels to the drones so that the time taken to deliver all the parcels is minimized. We prove that the natural approach of solving this problem via the LPT heuristic has an approximation factor of $\phi$, where $\phi \approx 1.62$ is the golden ratio., Comment: To appear in FSTTCS 2024

With the goal of building a concrete co-dimension one holographically dual field theory for four dimensional asymptotically flat spacetimes (4d AFS) as a limit of AdS$_4$/CFT$_3$, we begin an investigation of 3d Chern-Simons matter (CSM) theories in the Carroll regime. We perform a Carroll (speed of light $c\to0$) expansion of the relativistic Chern-Simons action coupled to a massless scalar and obtain Carrollian CSM theories, which we show are invariant under the infinite dimensional 3d conformal Carroll or 4d Bondi-van der Burg-Metzner-Sachs (BMS$_4$) symmetries, thus making them putative duals for 4d AFS. Concentrating on the leading-order electric Carroll CSM theory, we perform a null reduction of the 3d theory. Null reduction is a procedure to obtain non-relativistic theories from a higher dimensional relativistic theory. Curiously, null reduction of a Carrollian theory yields a relativistic lower-dimensional theory. We work with $SU(N) \times SU(M)$ CS theory coupled to bi-fundamental matter and show that when $N=M$, we obtain (rather surprisingly) a 2d Euclidean Yang-Mills theory after null reduction. We also comment on the reduction when $N \neq M$ and possible connections of the null-reduced Carroll theory to a candidate 2d Celestial CFT., Comment: 30 pages

Terahertz (THz) magnonics represent the notion of mathematical algebraic operations of magnons such as addition and subtraction in THz regime which is an emergent dissipationless ultrafast alternative to existing data processing technologies. Spin waves on antiferromagnets with a twist in spin order host such magnons in THz regime, which possess advantage of higher processing speeds, additional polarization degree of freedom and longer propagation lengths compared to that of gigahertz magnons in ferromagnets. While interaction among THz magnons is the crux of algebra operations, it requires magnetic orders with closely spaced magnon modes for easier experimental realization of their interactions. Herein, rich wealth of magnons spanning a narrow energy range of 0.4 to 10 meV is unraveled in Co4Ta2O9 using magneto-THz spectroscopy. Rare multitude of ten excitation modes, either of magnons or hybrid magnon-phonon modes is presented. Among other attributes, spin lattice interaction suggests a correlation among spin and local lattice distortion, magnetostriction, and magnetic exchange interaction signifying a THz magnetoelectric effect. This unification of structural, magnetic and dielectric facets, and their magnetic field control in a narrow spectrum unwinds the mechanism underneath the system's complexity while the manifestation of multitude of spin excitation modes is a potential source to design multiple channels in spin-wave computing based devices.

The complexity of interactions between the crystal-field and unusual non-collinear spin arrangement in non-trivial magnets demands novel tools to unravel the mystery underneath. In this work, we study such interaction dynamics of crystal-field-excitations (CFE) and low-energy magnetic excitations in orthochromite TmCrO3 with controls of temperature and magnetic field using high-resolution magneto-terahertz (THz) time-domain spectroscopy. The THz energy spectrum spanning 0.5-10 meV possesses a low-frequency spin-excitation (magnon) mode and a multitude of CFE modes at 10 K, all of which uniquely embody a range of phenomena. For the magnon mode, a temperature dependence of peak frequency is induced by magnetic interactions between Tm and Cr subsystems. While a change from blue- to red-shift of peak frequency of this mode marks the magnetization reversal transition, the spin reorientation temperature and change of magnetic anisotropy are depicted by different features of field- and temperature-dependent peak frequency dynamics. The modes corresponding to CFE are robust and laden with a multitude of sub-modes which are attributes of non-trivial interactions across different transitions. These modes are suppressed only upon substitution of Tb3+ at Tm3+ site, which suggests a dominant role of single-ion anisotropy in controlling entire THz excitations spectra. Overall, this remarkable range of phenomena seen through the unique lens of all-optical THz tools provides deeper insights into the origin of magnetic phases in systems with complex interactions between rare-earth and transition metal ions and provides a multitude of a novel combination of closely spaced modes for emerging hybrid spin-wave computation., Comment: The main copy of the manuscript includes 21 pages with 7 figures

Continual Test-Time Adaptation (CTTA) seeks to adapt a source pre-trained model to continually changing, unlabeled target domains. Existing TTA methods are typically designed for environments where domain changes occur sequentially and can struggle in more dynamic scenarios, as illustrated in Figure \ref{fig:settings}. Inspired by the principles of online K-Means, we introduce a novel approach to CTTA through visual prompting. We propose a \emph{Dynamic Prompt Coreset} that not only preserves knowledge from previously visited domains but also accommodates learning from new potential domains. This is complemented by a distance-based \emph{Weight Updating Mechanism} that ensures the coreset remains current and relevant. Our approach employs a fixed model architecture alongside the coreset and an innovative updating system to effectively mitigate challenges such as catastrophic forgetting and error accumulation. Extensive testing on four widely-used benchmarks demonstrates that our method consistently outperforms state-of-the-art alternatives in both classification and segmentation CTTA tasks across the structured and dynamic CTTA settings, with $99\%$ fewer trainable parameters.

Vision Transformers (ViTs) have demonstrated remarkable capabilities in learning representations, but their performance is compromised when applied to unseen domains. Previous methods either engage in prompt learning during the training phase or modify model parameters at test time through entropy minimization. The former often overlooks unlabeled target data, while the latter doesn't fully address domain shifts. In this work, our approach, Optimal Transport-guided Test-Time Visual Prompting (OT-VP), handles these problems by leveraging prompt learning at test time to align the target and source domains without accessing the training process or altering pre-trained model parameters. This method involves learning a universal visual prompt for the target domain by optimizing the Optimal Transport distance.OT-VP, with only four learned prompt tokens, exceeds state-of-the-art performance across three stylistic datasets-PACS, VLCS, OfficeHome, and one corrupted dataset ImageNet-C. Additionally, OT-VP operates efficiently, both in terms of memory and computation, and is adaptable for extension to online settings., Comment: WACV2025

Gauging the performance of ML models on data from unseen domains at test-time is essential yet a challenging problem due to the lack of labels in this setting. Moreover, the performance of these models on in-distribution data is a poor indicator of their performance on data from unseen domains. Thus, it is essential to develop metrics that can provide insights into the model's performance at test time and can be computed only with the information available at test time (such as their model parameters, the training data or its statistics, and the unlabeled test data). To this end, we propose a metric based on Optimal Transport that is highly correlated with the model's performance on unseen domains and is efficiently computable only using information available at test time. Concretely, our metric characterizes the model's performance on unseen domains using only a small amount of unlabeled data from these domains and data or statistics from the training (source) domain(s). Through extensive empirical evaluation using standard benchmark datasets, and their corruptions, we demonstrate the utility of our metric in estimating the model's performance in various practical applications. These include the problems of selecting the source data and architecture that leads to the best performance on data from an unseen domain and the problem of predicting a deployed model's performance at test time on unseen domains. Our empirical results show that our metric, which uses information from both the source and the unseen domain, is highly correlated with the model's performance, achieving a significantly better correlation than that obtained via the popular prediction entropy-based metric, which is computed solely using the data from the unseen domain.

The least squares method provides the best-fit curve by minimizing the total squares error. In this work, we provide the modified least squares method based on the fractional orthogonal polynomials that belong to the space $M_{n}^{\lambda} := \text{span}\{1,x^{\lambda},x^{2\lambda},\ldots,x^{n\lambda}\},~\lambda \in (0,2]$. Numerical experiments demonstrate how to solve different problems using the modified least squares method. Moreover, the results show the advantage of the modified least squares method compared to the classical least squares method. Furthermore, we discuss the various applications of the modified least squares method in the fields like fractional differential/integral equations and machine learning.

Estimating the out-of-distribution performance in regimes where labels are scarce is critical to safely deploy foundation models. Recently, it was shown that ensembles of neural networks observe the phenomena "agreement-on-the-line", which can be leveraged to reliably predict OOD performance without labels. However, in contrast to classical neural networks that are trained on in-distribution data from scratch for numerous epochs, foundation models undergo minimal finetuning from heavily pretrained weights, which may reduce the ensemble diversity needed to observe agreement-on-the-line. In our work, we demonstrate that when lightly finetuning multiple runs from a single foundation model, the choice of randomness during training (linear head initialization, data ordering, and data subsetting) can lead to drastically different levels of agreement-on-the-line in the resulting ensemble. Surprisingly, only random head initialization is able to reliably induce agreement-on-the-line in finetuned foundation models across vision and language benchmarks. Second, we demonstrate that ensembles of multiple foundation models pretrained on different datasets but finetuned on the same task can also show agreement-on-the-line. In total, by careful construction of a diverse ensemble, we can utilize agreement-on-the-line-based methods to predict the OOD performance of foundation models with high precision.

Interactions between Dark Matter (DM) and nucleons relevant for direct search experiments can be organised in a model independent manner using a Galiliean invariant, non--relativistic effective field theory (NREFT). Here one expands the interactions in powers of the momentum transfer $\vec{q}$ and DM velocity $\vec{v}$. This approach generates many operators. The potentially most important subleading operators are odd under $T$, and can thus only be present in a theory with $CP$ violating interactions. We consider two such operators, called $\mathcal{O}_{10}$ and $\mathcal{O}_{11}$ in the literature, in simplified models with neutral spin$-0$ mediators; the couplings are chosen such that the coefficient of the leading spin independent (SI) operator, which survives for $\vec{v} \rightarrow 0$, vanishes at tree level. However, it is generically induced at the next order in perturbation theory. We perform a numerical comparison of the number of scattering events between interactions involving the $T-$odd operators and the corresponding loop induced SI contributions. We find that for ''maximal'' $CP$ violation the former can dominate over the latter. However, in two of the three models we consider, an electric dipole moment of the neutron (nEDM) is induced at two--loop order. We find that the experimental bound on the nEDM typically leads to undetectably small rates induced by ${\mathcal O}_{10}$. On the other hand, the model leading to a nonvanishing coefficient of ${\mathcal O}_{11}$ does not induce an nEDM., Comment: 49 pages, 17 figures, 4 tables

In this paper, we investigate the potential of Large Language Models (LLMs) to improve English speaking skills. This is particularly relevant in countries like India, where English is crucial for academic, professional, and personal communication but remains a non-native language for many. Traditional methods for enhancing speaking skills often rely on human experts, which can be limited in terms of scalability, accessibility, and affordability. Recent advancements in Artificial Intelligence (AI) offer promising solutions to overcome these limitations. We propose Comuniqa, a novel LLM-based system designed to enhance English speaking skills. We adopt a human-centric evaluation approach, comparing Comuniqa with the feedback and instructions provided by human experts. In our evaluation, we divide the participants in three groups: those who use LLM-based system for improving speaking skills, those guided by human experts for the same task and those who utilize both the LLM-based system as well as the human experts. Using surveys, interviews, and actual study sessions, we provide a detailed perspective on the effectiveness of different learning modalities. Our preliminary findings suggest that while LLM-based systems have commendable accuracy, they lack human-level cognitive capabilities, both in terms of accuracy and empathy. Nevertheless, Comuniqa represents a significant step towards achieving Sustainable Development Goal 4: Quality Education by providing a valuable learning tool for individuals who may not have access to human experts for improving their speaking skills., Comment: Accepted at 7th ACM SIGCAS/SIGCHI Conference of Computing and Sustainable Societies : ACM COMPASS 2024

Wireless embedded edge devices are ubiquitous in our daily lives, enabling them to gather immense data via onboard sensors and mobile applications. This offers an amazing opportunity to train machine learning (ML) models in the realm of wireless devices for decision-making. Training ML models in a wireless setting necessitates transmitting datasets collected at the edge to a cloud parameter server, which is infeasible due to bandwidth constraints, security, and privacy issues. To tackle these challenges, Federated Learning (FL) has emerged as a distributed optimization approach to the decentralization of the model training process. In this work, we present a novel prototype to examine FL's effectiveness over bandwidth-constrained wireless channels. Through a novel design consisting of Zigbee and NI USRP devices, we propose a configuration that allows clients to broadcast synergistically local ML model updates to a central server to obtain a generalized global model. We assess the efficacy of this prototype using metrics such as global model accuracy and time complexity under varying conditions of transmission power, data heterogeneity and local learning., Comment: COMSNETS 2024 Demo Track (Accepted)

The epidemiological behavior of Plasmodium vivax malaria occurs across spatial scales including within-host, population, and metapopulation levels. On the within-host scale, P. vivax sporozoites inoculated in a host may form latent hypnozoites, the activation of which drives secondary infections and accounts for a large proportion of P. vivax illness; on the metapopulation level, the coupled human-vector dynamics characteristic of the population level are further complicated by the migration of human populations across patches with different malaria forces of (re-)infection. To explore the interplay of all three scales in a single two-patch model of Plasmodium vivax dynamics, we construct and study a system of eight integro-differential equations with periodic forcing (arising from the single-frequency sinusoidal movement of a human sub-population). Under the numerically-informed ansatz that the limiting solutions to the system are closely bounded by sinusoidal ones for certain regions of parameter space, we derive a single nonlinear equation from which all approximate limiting solutions may be drawn, and devise necessary and sufficient conditions for the equation to have only a disease-free solution. Our results illustrate the impact of movement on P. vivax transmission and suggest a need to focus vector control efforts on forest mosquito populations. The three-scale model introduced here provides a more comprehensive framework for studying the clinical, behavioral, and geographical factors underlying P. vivax malaria endemicity., Comment: 42 pages and 9 figures