Search

Contextual advertising serves ads that are aligned to the content that the user is viewing. The rapid growth of video content on social platforms and streaming services, along with privacy concerns, has increased the need for contextual advertising. Placing the right ad in the right context creates a seamless and pleasant ad viewing experience, resulting in higher audience engagement and, ultimately, better ad monetization. From a technology standpoint, effective contextual advertising requires a video retrieval system capable of understanding complex video content at a very granular level. Current text-to-video retrieval models based on joint multimodal training demand large datasets and computational resources, limiting their practicality and lacking the key functionalities required for ad ecosystem integration. We introduce ContextIQ, a multimodal expert-based video retrieval system designed specifically for contextual advertising. ContextIQ utilizes modality-specific experts-video, audio, transcript (captions), and metadata such as objects, actions, emotion, etc.-to create semantically rich video representations. We show that our system, without joint training, achieves better or comparable results to state-of-the-art models and commercial solutions on multiple text-to-video retrieval benchmarks. Our ablation studies highlight the benefits of leveraging multiple modalities for enhanced video retrieval accuracy instead of using a vision-language model alone. Furthermore, we show how video retrieval systems such as ContextIQ can be used for contextual advertising in an ad ecosystem while also addressing concerns related to brand safety and filtering inappropriate content., Comment: Accepted at WACV 2025

Quantum states of multiple qubits can violate Bell-type inequalities when there is entanglement present between the qubits, indicating nonlocal behaviour of correlations. We analyze the relation between multipartite entanglement and genuine multipartite nonlocality, characterized by Svetlichny inequality violations, for two families of $N-$qubit states. We show that for the generalized GHZ family of states, Svetlichny inequality is not violated when the $n-$tangle is less than $1/2$ for any number of qubits. On the other hand, the maximal slice states always violate the Svetlichny inequality when $n-$tangle is nonzero, and the violation increases monotonically with tangle when the number of qubits is even. Our work generalizes the relations between tangle and Svetlichny inequality violation previously derived for three qubits., Comment: 6 pages

Physics-informed neural networks (PINNs) are able to solve partial differential equations (PDEs) by incorporating the residuals of the PDEs into their loss functions. Variational Physics-Informed Neural Networks (VPINNs) and hp-VPINNs use the variational form of the PDE residuals in their loss function. Although hp-VPINNs have shown promise over traditional PINNs, they suffer from higher training times and lack a framework capable of handling complex geometries, which limits their application to more complex PDEs. As such, hp-VPINNs have not been applied in solving the Navier-Stokes equations, amongst other problems in CFD, thus far. FastVPINNs was introduced to address these challenges by incorporating tensor-based loss computations, significantly improving the training efficiency. Moreover, by using the bilinear transformation, the FastVPINNs framework was able to solve PDEs on complex geometries. In the present work, we extend the FastVPINNs framework to vector-valued problems, with a particular focus on solving the incompressible Navier-Stokes equations for two-dimensional forward and inverse problems, including problems such as the lid-driven cavity flow, the Kovasznay flow, and flow past a backward-facing step for Reynolds numbers up to 200. Our results demonstrate a 2x improvement in training time while maintaining the same order of accuracy compared to PINNs algorithms documented in the literature. We further showcase the framework's efficiency in solving inverse problems for the incompressible Navier-Stokes equations by accurately identifying the Reynolds number of the underlying flow. Additionally, the framework's ability to handle complex geometries highlights its potential for broader applications in computational fluid dynamics. This implementation opens new avenues for research on hp-VPINNs, potentially extending their applicability to more complex problems., Comment: 18 pages, 13 tables and 20 figures

Alzheimer's Disease (AD) is a neurodegenerative disease affecting millions of individuals across the globe. As the prevalence of this disease continues to rise, early diagnosis is crucial to improve clinical outcomes. Neural networks, specifically Convolutional Neural Networks (CNNs), are promising tools for diagnosing individuals with Alzheimer's. However, neural networks such as ANNs and CNNs typically yield lower validation accuracies when fed lower quantities of data. Hence, Generative Adversarial Networks (GANs) can be utilized to synthesize data to augment these existing MRI datasets, potentially yielding higher validation accuracies. In this study, we use this principle while examining a novel application of the SSMI metric in selecting high-quality synthetic data generated by our GAN to compare its accuracies with shuffled data generated by our GAN. We observed that incorporating GANs with an SSMI metric returned the highest accuracies when compared to a traditional dataset.

The first years of the 2000s led to an inflection point in computer architectures: while the number of available transistors on a chip continued to grow, crucial transistor scaling properties started to break down and result in increasing power consumption, while aggressive single-core performance optimizations were resulting in diminishing returns due to inherent limits in instruction-level parallelism. This led to the rise of multicore CPU architectures, which are now commonplace in modern computers at all scales. In this chapter, we discuss the evolution of multicore CPUs since their introduction. Starting with a historic overview of multiprocessing, we explore the basic microarchitecture of a multicore CPU, key challenges resulting from shared memory resources, operating system modifications to optimize multicore CPU support, popular metrics for multicore evaluation, and recent trends in multicore CPU design., Comment: preprint of book chapter to appear in Handbook of Computer Architecture

Classical chaos arises from the inherent non-linearity of dynamical systems. However, quantum maps are linear; therefore, the definition of chaos is not straightforward. To address this, we study a quantum system that exhibits chaotic behavior in its classical limit: the kicked top model, whose classical dynamics are governed by Hamilton's equations on phase space, whereas its quantum dynamics are described by the Schr\"odinger equation in Hilbert space. We explore the critical degree of non-linearity signifying the onset of chaos in the kicked top by modifying the original Hamiltonian so that the non-linearity is parametrized by a quantity $p$. We find two distinct behaviors of the modified kicked top depending on the value of $p$. Chaos intensifies as $p$ varies within the range of $1\leq p \leq 2$, whereas it diminishes for $p > 2$, eventually transitioning to a purely regular oscillating system as $p$ tends to infinity. We also comment on the complicated phase space structure for non-chaotic dynamics. Our investigation sheds light on the relationship between non-linearity and chaos in classical systems, offering insights into their dynamic behavior.

Point cloud understanding is an inherently challenging problem because of the sparse and unordered structure of the point cloud in the 3D space. Recently, Contrastive Vision-Language Pre-training (CLIP) based point cloud classification model i.e. PointCLIP has added a new direction in the point cloud classification research domain. In this method, at first multi-view depth maps are extracted from the point cloud and passed through the CLIP visual encoder. To transfer the 3D knowledge to the network, a small network called an adapter is fine-tuned on top of the CLIP visual encoder. PointCLIP has two limitations. Firstly, the point cloud depth maps lack image information which is essential for tasks like classification and recognition. Secondly, the adapter only relies on the global representation of the multi-view features. Motivated by this observation, we propose a Pretrained Point Cloud to Image Translation Network (PPCITNet) that produces generalized colored images along with additional salient visual cues to the point cloud depth maps so that it can achieve promising performance on point cloud classification and understanding. In addition, we propose a novel viewpoint adapter that combines the view feature processed by each viewpoint as well as the global intertwined knowledge that exists across the multi-view features. The experimental results demonstrate the superior performance of the proposed model over existing state-of-the-art CLIP-based models on ModelNet10, ModelNet40, and ScanobjectNN datasets., Comment: Accepted by ICPR2024

Despite the prevalent view that quantum mechanics is irrelevant to macroscopic biological systems because of inherent noise and decoherence, this paper demonstrates that the electrical noise (Brownian motion) in neuron membranes gives rise to an `emergent' Schr\"{o}dinger equation involving a new neuronal constant $\hat{\hbar}$, fundamentally challenging the standard view of neuronal behaviour. This result could provide new insights into the underlying mechanisms of brain function, thus challenging existing paradigms in both quantum physics and neuroscience. A possible empirical test of this emergent quantum behaviour would be to look for quantum fluctuations in subthreshold neural oscillations., Comment: 12 pages, 5 figures; title and abstract revised; manuscript expanded with explanatory statements and figures and some new results, keeping the original claim intact

Stochastic video generation is particularly challenging when the camera is mounted on a moving platform, as camera motion interacts with observed image pixels, creating complex spatio-temporal dynamics and making the problem partially observable. Existing methods typically address this by focusing on raw pixel-level image reconstruction without explicitly modelling camera motion dynamics. We propose a solution by considering camera motion or action as part of the observed image state, modelling both image and action within a multi-modal learning framework. We introduce three models: Video Generation with Learning Action Prior (VG-LeAP) treats the image-action pair as an augmented state generated from a single latent stochastic process and uses variational inference to learn the image-action latent prior; Causal-LeAP, which establishes a causal relationship between action and the observed image frame at time $t$, learning an action prior conditioned on the observed image states; and RAFI, which integrates the augmented image-action state concept into flow matching with diffusion generative processes, demonstrating that this action-conditioned image generation concept can be extended to other diffusion-based models. We emphasize the importance of multi-modal training in partially observable video generation problems through detailed empirical studies on our new video action dataset, RoAM.

Realization of the stationary integral solutions of steady state transonic accretion flow in spherical symmetry helps to understand accretion phenomena on various astrophysical objects. In recent years, attempts have been made to study accreting black hole systems as an example of autonomous dynamical systems. The fixed point analysis is used to study the transonic properties of accretion flow onto an astrophysical black hole, hence the nature of the phase orbits for the transonic flow solutions can be understood without constructing the integral solutions. Since a large-scale astrophysical fluid flow is vulnerable to external perturbation, one needs to ensure that the stationary accretion solutions are stable under perturbation. By adopting a time-dependent stability analysis scheme for the accretion flow, one demonstrates under which condition the perturbation will not diverge. It has also been observed that a space-time metric, dubbed the sonic metric, can be constructed to describe the propagation of the perturbation embedded within the accreting fluid, which mimics a black hole like spacetime within the accreting fluid, where the transonic surfaces can be identified with a black hole like horizons. Such identification is accomplished using the theory of causal structure - by constructing Carter-Penrose diagrams. An accreting black hole system, thus, can be perceived as a classical analogue gravity model naturally found in the universe. Hence, accretion phenomena onto astrophysical black holes can be looked upon from three apparently non-overlapping perspectives - astrophysical processes, theory of dynamical systems, and emergent gravity (alternatively, the analogue gravity) phenomena, respectively. The present article illustrates, by taking the simplest possible accretion flow model, how one can study astrophysical accretion processes from the aforementioned perspectives (Abridged)., Comment: 10 pages, 3 figues, RevTex Class, Comments Welcome

Existing UAS Traffic Management (UTM) frameworks designate preplanned flight paths to uncrewed aircraft systems (UAS), enabling the UAS to deliver payloads. However, with increasing delivery demand between the source-destination pairs in the urban airspace, UAS will likely experience considerable congestion on the nominal paths. We propose a rule-based congestion mitigation strategy that improves UAS safety and airspace utilization in congested traffic streams. The strategy relies on nominal path information from the UTM and positional information of other UAS in the vicinity. Following the strategy, UAS opts for alternative local paths in the unoccupied airspace surrounding the nominal path and avoids congested regions. The strategy results in UAS traffic exploring and spreading to alternative adjacent routes on encountering congestion. The paper presents queuing models to estimate the expected traffic spread for varying stochastic delivery demand at the source, thus helping to reserve the airspace around the nominal path beforehand to accommodate any foreseen congestion. Simulations are presented to validate the queuing results in the presence of static obstacles and intersecting UAS streams.

Boolean satisfiability (SAT) problems are routinely solved by SAT solvers in real-life applications, yet solving time can vary drastically between solvers for the same instance. This has motivated research into machine learning models that can predict, for a given SAT instance, which solver to select among several options. Existing SAT solver selection methods all rely on some hand-picked instance features, which are costly to compute and ignore the structural information in SAT graphs. In this paper we present GraSS, a novel approach for automatic SAT solver selection based on tripartite graph representations of instances and a heterogeneous graph neural network (GNN) model. While GNNs have been previously adopted in other SAT-related tasks, they do not incorporate any domain-specific knowledge and ignore the runtime variation introduced by different clause orders. We enrich the graph representation with domain-specific decisions, such as novel node feature design, positional encodings for clauses in the graph, a GNN architecture tailored to our tripartite graphs and a runtime-sensitive loss function. Through extensive experiments, we demonstrate that this combination of raw representations and domain-specific choices leads to improvements in runtime for a pool of seven state-of-the-art solvers on both an industrial circuit design benchmark, and on instances from the 20-year Anniversary Track of the 2022 SAT Competition., Comment: Accepted by KDD 2024

As more distributed energy resources become part of the demand-side infrastructure, it is important to quantify the energy flexibility they provide on a community scale, particularly to understand the impact of geographic, climatic, and occupant behavioral differences on their effectiveness, as well as identify the best control strategies to accelerate their real-world adoption. CityLearn provides an environment for benchmarking simple and advanced distributed energy resource control algorithms including rule-based, model-predictive, and reinforcement learning control. CityLearn v2 presented here extends CityLearn v1 by providing a simulation environment that leverages the End-Use Load Profiles for the U.S. Building Stock dataset to create virtual grid-interactive communities for resilient, multi-agent distributed energy resources and objective control with dynamic occupant feedback. This work details the v2 environment design and provides application examples that utilize reinforcement learning to manage battery energy storage system charging/discharging cycles, vehicle-to-grid control, and thermal comfort during heat pump power modulation.

Variational Physics-Informed Neural Networks (VPINNs) utilize a variational loss function to solve partial differential equations, mirroring Finite Element Analysis techniques. Traditional hp-VPINNs, while effective for high-frequency problems, are computationally intensive and scale poorly with increasing element counts, limiting their use in complex geometries. This work introduces FastVPINNs, a tensor-based advancement that significantly reduces computational overhead and improves scalability. Using optimized tensor operations, FastVPINNs achieve a 100-fold reduction in the median training time per epoch compared to traditional hp-VPINNs. With proper choice of hyperparameters, FastVPINNs surpass conventional PINNs in both speed and accuracy, especially in problems with high-frequency solutions. Demonstrated effectiveness in solving inverse problems on complex domains underscores FastVPINNs' potential for widespread application in scientific and engineering challenges, opening new avenues for practical implementations in scientific machine learning., Comment: 31 pages, 19 figures, 4 algorithms

Intelligent energy management strategies, such as Vehicle-to-Grid (V2G) and Grid-to-Vehicle (G2V) emerge as a potential solution to the Electric Vehicles' (EVs) integration into the energy grid. These strategies promise enhanced grid resilience and economic benefits for both vehicle owners and grid operators. Despite the announced prospective, the adoption of these strategies is still hindered by an array of operational problems. Key among these is the lack of a simulation platform that allows to validate and refine V2G and G2V strategies. Including the development, training, and testing in the context of Energy Communities (ECs) incorporating multiple flexible energy assets. Addressing this gap, first we introduce the EVLearn, a simulation module for researching in both V2G and G2V energy management strategies, that models EVs, their charging infrastructure and associated energy flexibility dynamics; second, this paper integrates EVLearn with the existing CityLearn framework, providing V2G and G2V simulation capabilities into the study of broader energy management strategies. Results validated EVLearn and its integration into CityLearn, where the impact of these strategies is highlighted through a comparative simulation scenario., Comment: 10 pages, 7 figures, 3 tables, 11 equations

Long-term video generation and prediction remain challenging tasks in computer vision, particularly in partially observable scenarios where cameras are mounted on moving platforms. The interaction between observed image frames and the motion of the recording agent introduces additional complexities. To address these issues, we introduce the Action-Conditioned Video Generation (ACVG) framework, a novel approach that investigates the relationship between actions and generated image frames through a deep dual Generator-Actor architecture. ACVG generates video sequences conditioned on the actions of robots, enabling exploration and analysis of how vision and action mutually influence one another in dynamic environments. We evaluate the framework's effectiveness on an indoor robot motion dataset which consists of sequences of image frames along with the sequences of actions taken by the robotic agent, conducting a comprehensive empirical study comparing ACVG to other state-of-the-art frameworks along with a detailed ablation study.

Point cloud classification refers to the process of assigning semantic labels or categories to individual points within a point cloud data structure. Recent works have explored the extension of pre-trained CLIP to 3D recognition. In this direction, CLIP-based point cloud models like PointCLIP, CLIP2Point have become state-of-the-art methods in the few-shot setup. Although these methods show promising performance for some classes like airplanes, desks, guitars, etc, the performance for some classes like the cup, flower pot, sink, nightstand, etc is still far from satisfactory. This is due to the fact that the adapter of CLIP-based models is trained using randomly sampled N-way K-shot data in the standard supervised learning setup. In this paper, we propose a novel meta-episodic learning framework for CLIP-based point cloud classification, addressing the challenges of limited training examples and sampling unknown classes. Additionally, we introduce dynamic task sampling within the episode based on performance memory. This sampling strategy effectively addresses the challenge of sampling unknown classes, ensuring that the model learns from a diverse range of classes and promotes the exploration of underrepresented categories. By dynamically updating the performance memory, we adaptively prioritize the sampling of classes based on their performance, enhancing the model's ability to handle challenging and real-world scenarios. Experiments show an average performance gain of 3-6\% on ModelNet40 and ScanobjectNN datasets in a few-shot setup.

Active learning (AL) techniques optimally utilize a labeling budget by iteratively selecting instances that are most valuable for learning. However, they lack ``prerequisite checks'', i.e., there are no prescribed criteria to pick an AL algorithm best suited for a dataset. A practitioner must pick a technique they \emph{trust} would beat random sampling, based on prior reported results, and hope that it is resilient to the many variables in their environment: dataset, labeling budget and prediction pipelines. The important questions then are: how often on average, do we expect any AL technique to reliably beat the computationally cheap and easy-to-implement strategy of random sampling? Does it at least make sense to use AL in an ``Always ON'' mode in a prediction pipeline, so that while it might not always help, it never under-performs random sampling? How much of a role does the prediction pipeline play in AL's success? We examine these questions in detail for the task of text classification using pre-trained representations, which are ubiquitous today. Our primary contribution here is a rigorous evaluation of AL techniques, old and new, across setups that vary wrt datasets, text representations and classifiers. This unlocks multiple insights around warm-up times, i.e., number of labels before gains from AL are seen, viability of an ``Always ON'' mode and the relative significance of different factors. Additionally, we release a framework for rigorous benchmarking of AL techniques for text classification.

As the amount of data produced in society continues to grow at an exponential rate, modern applications are incurring significant performance and energy penalties due to high data movement between the CPU and memory/storage. While processing in main memory can alleviate these penalties, it is becoming increasingly difficult to keep large datasets entirely in main memory. This has led to a recent push for in-storage computation, where processing is performed inside the storage device. We propose TCAM-SSD, a new framework for search-based computation inside the NAND flash memory arrays of a conventional solid-state drive (SSD), which requires lightweight modifications to only the array periphery and firmware. TCAM-SSD introduces a search manager and link table, which can logically partition the NAND flash memory's contents into search-enabled regions and standard storage regions. Together, these light firmware changes enable TCAM-SSD to seamlessly handle block I/O operations, in addition to new search operations, thereby reducing end-to-end execution time and total data movement. We provide an NVMe-compatible interface that provides programmers with the ability to dynamically allocate data on and make use of TCAM-SSD, allowing the system to be leveraged by a wide variety of applications. We evaluate three example use cases of TCAM-SSD to demonstrate its benefits. For transactional databases, TCAM-SSD can mitigate the performance penalties for applications with large datasets, achieving a 60.9% speedup over a conventional system that retrieves data from the SSD and computes using the CPU. For database analytics, TCAM-SSD provides an average speedup of 17.7x over a conventional system for a collection of analytical queries. For graph analytics, we combine TCAM-SSD's associative search with a sparse data structure, speeding up graph computing for larger-than-memory datasets by 14.5%.

The lack of diversity in physics remains a persistent worldwide problem. Despite being a quantitative discipline which relies on measurements to construct and validate hypotheses, there remains a paucity of data on both demographics and experiences of marginalized groups. In Canada, there has never been a nationwide assessment of those studying or working in physics. Here, we present findings from Canadian Physics Counts: the first national survey of equity, diversity, and inclusion (EDI) in the Canadian physics community. Our intersectional approach allowed us to gather a wealth of information on gender identity, sexual orientation, race, disability, and more. Analyses revealed key findings, including the first data on physicists who identify as non-binary or gender diverse, as well as the first data on Black and Indigenous scholars. Black physicists (1.2%) and Indigenous physicists (.3%) were found to be the most underrepresented, while White men were overrepresented across all sectors. Among respondents with a disability, 5% reported receiving full accommodations for their required needs at their place of work or study. One in four respondents from BIPOC gender diverse backgrounds identified as being disabled, and the proportion of sexually diverse students who reported having a disability was more than three times higher than the proportion of heterosexual students with a disability. The data also revealed that students represented more demographic diversity than working professionals, highlighting the importance of acting today in order to retain the diverse physicists of tomorrow. Our analysis identifies areas for intervention and offers recommendations for building a diverse and inclusive physics community in Canada that can be a global exemplar., Comment: 26 pages, 7 figures

In this paper, we provide expressions for the secrecy outage probability (SOP) for suboptimal and optimal opportunistic scheduling schemes in a reconfigurable intelligent surface (RIS) aided system with multiple eavesdroppers in approximate closed form. A suboptimal scheduling (SS) scheme is analyzed, which is used when the channel state information (CSI) of the eavesdropping links is unavailable, and the optimal scheduling (OS) scheme is also analyzed, which is used when the global CSI is available. For each scheme, we provide a simplified expression for the SOP in the high signal-to-noise ratio (SNR) regime to demonstrate its behavior as a function of the key system parameters. At high SNR, the SOP saturates to a constant level which decreases exponentially with the number of RIS elements in the SS scheme and with the product of the number of RIS elements and the number of users in the OS scheme. We compare the performance of the opportunistic user scheduling schemes with that of a non-orthogonal multiple access (NOMA) based scheduling scheme which chooses a pair of users in each time slot for scheduling and we show that the opportunistic schemes outperform the NOMA-based scheme. We also derive a closed-form expression for the SOP of a decode-and-forward (DF) relay-aided scheduling scheme in order to compare it with that of the RIS-aided system. It is found that the RIS-aided system outperforms the relay-aided systems when the number of RIS elements is sufficiently large. An increased number of RIS elements is required to outperform the relay-aided system at higher operating frequencies.

Artificial intelligence is seen as increasingly important, and potentially profoundly so, but the fields of AI ethics and AI engineering have not fully recognized that these technologies, including large language models (LLMs), will have massive impacts on animals. We argue that this impact matters, because animals matter morally. As a first experiment in evaluating animal consideration in LLMs, we constructed a proof-of-concept Evaluation System, which assesses LLM responses and biases from multiple perspectives. This system evaluates LLM outputs by two criteria: their truthfulness, and the degree of consideration they give to the interests of animals. We tested OpenAI ChatGPT 4 and Anthropic Claude 2.1 using a set of structured queries and predefined normative perspectives. Preliminary results suggest that the outcomes of the tested models can be benchmarked regarding the consideration they give to animals, and that generated positions and biases might be addressed and mitigated with more developed and validated systems. Our research contributes one possible approach to integrating animal ethics in AI, opening pathways for future studies and practical applications in various fields, including education, public policy, and regulation, that involve or relate to animals and society. Overall, this study serves as a step towards more useful and responsible AI systems that better recognize and respect the vital interests and perspectives of all sentient beings., Comment: AAAI 2024 Workshop on Public Sector LLMs: Algorithmic and Sociotechnical Design. 12 pages, 11 figures

Processing-using-DRAM (PUD) is a processing-in-memory (PIM) approach that uses a DRAM array's massive internal parallelism to execute very-wide data-parallel operations, in a single-instruction multiple-data (SIMD) fashion. However, DRAM rows' large and rigid granularity limit the effectiveness and applicability of PUD in three ways. First, since applications have varying degrees of SIMD parallelism, PUD execution often leads to underutilization, throughput loss, and energy waste. Second, most PUD architectures are limited to the execution of parallel map operations. Third, the need to feed the wide DRAM row with tens of thousands of data elements combined with the lack of adequate compiler support for PUD systems create a programmability barrier. Our goal is to design a flexible PUD system that overcomes the limitations caused by the large and rigid granularity of PUD. To this end, we propose MIMDRAM, a hardware/software co-designed PUD system that introduces new mechanisms to allocate and control only the necessary resources for a given PUD operation. The key idea of MIMDRAM is to leverage fine-grained DRAM (i.e., the ability to independently access smaller segments of a large DRAM row) for PUD computation. MIMDRAM exploits this key idea to enable a multiple-instruction multiple-data (MIMD) execution model in each DRAM subarray. We evaluate MIMDRAM using twelve real-world applications and 495 multi-programmed application mixes. Our evaluation shows that MIMDRAM provides 34x the performance, 14.3x the energy efficiency, 1.7x the throughput, and 1.3x the fairness of a state-of-the-art PUD framework, along with 30.6x and 6.8x the energy efficiency of a high-end CPU and GPU, respectively. MIMDRAM adds small area cost to a DRAM chip (1.11%) and CPU die (0.6%)., Comment: Extended version of HPCA 2024 paper. arXiv admin note: text overlap with arXiv:2109.05881 by other authors

The entry of large language models (LLMs) into research and commercial spaces has led to a trend of ever-larger models, with initial promises of generalisability, followed by a widespread desire to downsize and create specialised models without the need for complete fine-tuning, using Parameter Efficient Fine-tuning (PEFT) methods. We present an investigation into the suitability of different PEFT methods to clinical decision-making tasks, across a range of model sizes, including extremely small models with as few as $25$ million parameters. Our analysis shows that the performance of most PEFT approaches varies significantly from one task to another, with the exception of LoRA, which maintains relatively high performance across all model sizes and tasks, typically approaching or matching full fine-tuned performance. The effectiveness of PEFT methods in the clinical domain is evident, particularly for specialised models which can operate on low-cost, in-house computing infrastructure. The advantages of these models, in terms of speed and reduced training costs, dramatically outweighs any performance gain from large foundation LLMs. Furthermore, we highlight how domain-specific pre-training interacts with PEFT methods and model size, and discuss how these factors interplay to provide the best efficiency-performance trade-off. Full code available at: tbd.

A method is proposed to produce a classical optical state that is `intersystem nonseparable' and a close analog of the $\phi^+$ Bell state. A derivation of the CHSH-Bell inequality is sketched within the framework of classical polarization optics, using non-contextuality for factorizable states as an axiom rather than any hidden variable theory. It is shown that the classical state violates this inequality., Comment: 10 pages, two figures (Abstract slightly modified); typos corrected, stylistic changes made, one figure added, one Appendix added, title slightly modified; no fundamental changes made

In this paper, we study underlay device-to-device (D2D) communication systems empowered by a reconfigurable intelligent surface (RIS) for cognitive cellular networks. Considering Rayleigh fading channels and the general case where there exist both the direct and RIS-enabled D2D channels, the outage probability (OP) of the D2D communication link is presented in closed-form. Next, for the considered RIS-empowered underlaid D2D system, we frame an OP minimization problem. We target the joint optimization of the transmit power at the D2D source and the RIS placement, under constraints on the transmit power at the D2D source and on the limited interference imposed on the cellular user for two RIS deployment topologies. Due to the coupled optimization variables, the formulated optimization problem is extremely intractable. We propose an equivalent transformation which we are able to solve analytically. In the transformed problem, an expression for the average value of the signal-to-interference-noise ratio (SINR) at the D2D receiver is derived in closed-form. Our theoretical derivations are corroborated through simulation results, and various system design insights are deduced. It is indicatively showcased that the proposed RIS-empowered underlaid D2D system design outperforms the benchmark semi-adaptive optimal power and optimal distance schemes, offering $44\%$ and $20\%$ performance improvement, respectively.

In this paper, we address the problem of predicting collision for objects moving on the surface of a spherical manifold. Toward this end, we develop the notion of a collision triangle on such manifolds. We use this to determine analytical conditions governing the speed ratios and direcions of motion of objects that lead to collisions on the sphere. We first develop these conditions for point objects and subsequently extend this to the case of circular patch shaped objects moving on the sphere., Comment: 14 pages

Methods: Broad- and narrow-band imaging of AFGL 5180 was made in the NIR with the LBT, in both seeing-limited ($\sim0.5\arcsec$) and high angular resolution ($\sim0.09\arcsec$) Adaptive Optics (AO) modes, as well as with HST. Archival ALMA continuum data was also utilized. Results: At least 40 jet knots were identified via NIR emission from H$_2$ and [FeII] tracing shocked gas. Bright jet knots outflowing from the central most massive protostar, S4, are detected towards the east of the source and are resolved in fine detail with the AO imaging. Additional knots are distributed throughout the field, likely indicating the presence of multiple driving sources. Sub-millimeter sources detected by ALMA are shown to be grouped in two main complexes, AFGL 5180 M and a small cluster $\sim15\arcsec$ to the south, AFGL 5180 S. From our NIR continuum images we identify YSO candidates down to masses of $\sim 0.1\:M_\odot$. Combined with the sub-mm sources, this yields a surface number density of such YSOs of $N_* \sim 10^3 {\rm pc}^{-2}$ within a projected radius of about 0.1 pc. Such a value is similar to those predicted by models of both Core Accretion from a turbulent clump environment and Competitive Accretion. The radial profile of $N_*$ is relatively flat on scales out to 0.2~pc, with only modest enhancement around the massive protostar inside 0.05~pc. Conclusions: This study demonstrates the utility of high-resolution NIR imaging, in particular with AO, for detecting outflow activity and YSOs in distant regions. The presented images reveal the complex morphology of outflow-shocked gas within the large-scale bipolar flow of a massive protostar, as well as clear evidence for several other outflow driving sources in the region. Finally, this work presents a novel approach to compare the observed YSO surface number density from our study against different models of massive star formation., Comment: Accepted to Astronomy & Astrophysics (A&A)

The SOUL systems at the Large Bincoular Telescope can be seen such as precursor for the ELT SCAO systems, combining together key technologies such as EMCCD, Pyramid WFS and adaptive telescopes. After the first light of the first upgraded system on September 2018, going through COVID and technical stops, we now have all the 4 systems working on-sky. Here, we report about some key control improvements and the system performance characterized during the commissioning. The upgrade allows us to correct more modes (500) in the bright end and increases the sky coverage providing SR(K)>20% with reference stars G$_{RP}$<17, opening to extragalcatic targets with NGS systems. Finally, we review the first astrophysical results, looking forward to the next generation instruments (SHARK-NIR, SHARK-Vis and iLocater), to be fed by the SOUL AO correction., Comment: 13 pages, 10 figures, Adaptive Optics for Extremely Large Telescopes 7th Edition, 25-30 Jun 2023 Avignon (France)

The bio-acoustic information contained within heart sound signals are utilized by physicians world-wide for auscultation purpose. However, the heart sounds are inherently susceptible to noise contamination. Various sources of noises like lung sound, coughing, sneezing, and other background noises are involved in such contamination. Such corruption of the heart sound signal often leads to inconclusive or false diagnosis. To address this issue, we have proposed a novel U-Net based deep neural network architecture for denoising of phonocardiogram (PCG) signal in this paper. For the design, development and validation of the proposed architecture, a novel approach of synthesizing real-world noise corrupted PCG signals have been proposed. For the purpose, an open-access real-world noise sample dataset and an open-access PCG dataset has been utilized. The performance of the proposed denoising methodology has been evaluated on the synthesized noisy PCG dataset. The performance of the proposed algorithm has been compared with existing state-of-the-art (SoA) denoising algorithms qualitatively and quantitatively. The proposed denoising technique has shown improvement in performance as comparison to the SoAs.