Search

We construct a finite deterministic graphical (DG) game without Nash equilibria in pure stationary strategies. This game has 3 players $I=\{1,2,3\}$ and 5 outcomes: 2 terminal $a_1$ and $a_2$ and 3 cyclic. Furthermore, for 2 players a terminal outcome is the best: $a_1$ for player 3 and $a_2$ for player 1. Hence, the rank vector $r$ is at most $(1,2,1)$. Here $r_i$ is the number of terminal outcomes that are worse than some cyclic outcome for the player $i \in I$. This is a counterexample to conjecture ``Catch 22" from the paper ``On Nash-solvability of finite $n$-person DG games, Catch 22" (2021) arXiv:2111.06278, according to which, at least 2 entries of $r$ are at least 2 for any NE-free game. However, Catch 22 remains still open for the games with a unique cyclic outcome, not to mention a weaker (and more important) conjecture claiming that an $n$-person finite DG game has a Nash equilibrium (in pure stationary strategies) when $r = (0^n)$, that is, all $n$ entries of $r$ are 0; in other words, when the following condition holds: $\qquad\bullet$ ($C_0$) any terminal outcome is better than every cyclic one for each player. A game is play-once if each player controls a unique position. It is known that any play-once game satisfying ($C_0$) has a Nash equilibrium. We give a new and very short proof of this statement. Yet, not only conjunction but already disjunction of the above two conditions may be sufficient for Nash-solvability. This is still open.

The anti-windup compensation typically addresses strict control limitations in control systems. However, there is a clear need for an equivalent solution for the states/outputs of the system. This paper introduces a novel methodology for the state anti-windup compensator. Unlike state-constrained control methods, which often focus on incorporating soft constraints into the design or fail to react adequately to constraint violations in practical settings, the proposed methodology treats state constraints as implement-oriented soft-hard constraints. This is achieved by integrating a saturation block within the structure of the safety compensator, referred to as the state anti-windup (SANTW) compensator. Similar to input anti-windup schemes, the SANTW design is separated from the nominal controller design. The problem is formulated as a disturbance rejection one to directly minimize the saturation. The paper develops two Hinf optimization frameworks using frequency-domain solutions and linear matrix inequalities. It then addresses constraints on both inputs and states, resulting in a unified Input-State Anti-windup (IS-ANTW) compensator synthesized using non-smooth Hinf optimization. This method also offers the flexibility of having a fixed-order compensator, crucial in many practical applications. Additionally, the study evaluates the proposed compensator's performance in managing current fluctuations from renewable energy sources during grid faults, demonstrating its effectiveness through detailed Electromagnetic Transient (EMT) simulations of grid-connected DC-AC converters., Comment: 15 pages 21 Figures

We consider a ubiquitous scenario in the study of Influence Maximization (IM), in which there is limited knowledge about the topology of the diffusion network. We set the IM problem in a multi-round diffusion campaign, aiming to maximize the number of distinct users that are influenced. Leveraging the capability of bandit algorithms to effectively balance the objectives of exploration and exploitation, as well as the expressivity of neural networks, our study explores the application of neural bandit algorithms to the IM problem. We propose the framework IM-GNB (Influence Maximization with Graph Neural Bandits), where we provide an estimate of the users' probabilities of being influenced by influencers (also known as diffusion seeds). This initial estimate forms the basis for constructing both an exploitation graph and an exploration one. Subsequently, IM-GNB handles the exploration-exploitation tradeoff, by selecting seed nodes in real-time using Graph Convolutional Networks (GCN), in which the pre-estimated graphs are employed to refine the influencers' estimated rewards in each contextual setting. Through extensive experiments on two large real-world datasets, we demonstrate the effectiveness of IM-GNB compared with other baseline methods, significantly improving the spread outcome of such diffusion campaigns, when the underlying network is unknown., Comment: To appear at the 2024 ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD)

The term of big data was used since 1990s, but it became very popular around 2012. A recent definition of this term says that big data are information assets characterized by high volume, velocity, variety and veracity that need special analytical methods and software technologies to extract value form them. While big data was used at the beginning mostly in information technology field, now it can be found in every area of activity: in governmental decision-making processes, manufacturing, education, healthcare, economics, engineering, natural sciences, sociology. The rise of Internet, mobile phones, social media networks, different types of sensors or satellites provide enormous quantities of data that can have profound effects on economic research. The data revolution that we are facing transformed the way we measure the human behavior and economic activities. Unemployment, consumer price index, population mobility, financial transactions are only few examples of economic phenomena that can be analyzed using big data sources. In this paper we will start with a taxonomy of big data sources and show how these new data sources can be used in empirical analyses and to build economic indicators very fast and with reduced costs., Comment: Knowledge for the market use, Conference Proceedings, 2018, Olomouc, Czech Republic

Transformers and LLMs have seen rapid adoption in all domains. Their sizes have exploded to hundreds of billions of parameters and keep increasing. Under these circumstances, the training of transformers is slow and often takes in the order of weeks or months. Thanks to 3D model parallelism (data, pipeline, and tensor-level parallelism), the training can scale to a large number of GPUs, which reduces the duration of the training but dramatically increases the cost. Even when a large number of GPUs are available, the aggregated GPU memory is often not enough to hold the full training state (optimizer state, model parameters, and gradients). To compensate, state-of-the-art approaches offload the optimizer state at least partially to the host memory and perform hybrid CPU-GPU computations. Such flexible solutions dramatically reduce the GPU memory utilization, which makes it feasible to run the training on a smaller number of GPUs at the cost of performance penalty. Unfortunately, the challenges and bottlenecks of adopting this strategy are not sufficiently studied by state-of-the-art, which results in poor management of the combined host-GPU memory and poor overlapping between data movements and computations. In this paper, we aim to fill this gap by characterizing the behavior of offloaded training using the DeepSpeed runtime. Specifically, we study the GPU memory utilization over time during each iteration, the activity on the PCIe related to transfers between the host memory and the GPU memory, and the relationship between resource utilization and the steps involved in each iteration. Thanks to this study, we reveal opportunities for future improvements of offloading solutions, which enable greater flexibility to optimize the cost-performance trade-off in the context of transformer and LLM training., Comment: Accepted at FlexScience'24' Workshop on AI and Scientific Computing at Scale using Flexible Computing Infrastructures (co-located with HPDC'24)

LLMs have seen rapid adoption in all domains. They need to be trained on high-end high-performance computing (HPC) infrastructures and ingest massive amounts of input data. Unsurprisingly, at such a large scale, unexpected events (e.g., failures of components, instability of the software, undesirable learning patterns, etc.), are frequent and typically impact the training in a negative fashion. Thus, LLMs need to be checkpointed frequently so that they can be rolled back to a stable state and subsequently fine-tuned. However, given the large sizes of LLMs, a straightforward checkpointing solution that directly writes the model parameters and optimizer state to persistent storage (e.g., a parallel file system), incurs significant I/O overheads. To address this challenge, in this paper we study how to reduce the I/O overheads for enabling fast and scalable checkpointing for LLMs that can be applied at high frequency (up to the granularity of individual iterations) without significant impact on the training process. Specifically, we introduce a lazy asynchronous multi-level approach that takes advantage of the fact that the tensors making up the model and optimizer state shards remain immutable for extended periods of time, which makes it possible to copy their content in the background with minimal interference during the training process. We evaluate our approach at scales of up to 180 GPUs using different model sizes, parallelism settings, and checkpointing frequencies. The results show up to 48$\times$ faster checkpointing and 2.2$\times$ faster end-to-end training runtime compared with the state-of-art checkpointing approaches., Comment: Published at HPDC '24: The 33rd International Symposium on High-Performance Parallel and Distributed Computing. Source code at https://github.com/DataStates/datastates-llm

Topologically trivial insulators can be classified into atomic insulators (AIs) and obstructed atomic insulators (OAIs) depending on whether the Wannier charge centers are localized or not at spatial positions occupied by atoms. An OAI can possess unusual properties such as surface states along certain crystalline surfaces, which advantageously appear in materials with much larger bulk energy gap than topological insulators, making them more attractive for potential applications. In this work, we show that a well-known crystal, silicon (Si) is a model OAI, which naturally explains some of Si's unusual properties such as its famous (111) surface states. On this surface, using angle resolved photoemission spectroscopy (ARPES), we reveal sharp quasi-1D massive Dirac line dispersions; we also observe, using scanning tunneling microscopy/spectroscopy (STM/STS), topological solitons at the interface of the two atomic chains. Remarkably, we show that the different chain domains can be reversibly switched at the nanometer scale, suggesting the application potential in ultra-high density storage devices.

Multimodal Large Language Models (MLLMs) have demonstrated a wide range of capabilities across many domains, including Embodied AI. In this work, we study how to best ground a MLLM into different embodiments and their associated action spaces, with the goal of leveraging the multimodal world knowledge of the MLLM. We first generalize a number of methods through a unified architecture and the lens of action space adaptors. For continuous actions, we show that a learned tokenization allows for sufficient modeling precision, yielding the best performance on downstream tasks. For discrete actions, we demonstrate that semantically aligning these actions with the native output token space of the MLLM leads to the strongest performance. We arrive at these lessons via a thorough study of seven action space adapters on five different environments, encompassing over 114 embodied tasks.

In this paper, we study cascading failures in power grids through the lens of information diffusion models. Similar to the spread of rumors or influence in an online social network, it has been observed that failures (outages) in a power grid can spread contagiously, driven by viral spread mechanisms. We employ a stochastic diffusion model that is Markovian (memoryless) and local (the activation of one node, i.e., transmission line, can only be caused by its neighbors). Our model integrates viral diffusion principles with physics-based concepts, by correlating the diffusion weights (contagion probabilities between transmission lines) with the hyperparametric Information Cascades (IC) model. We show that this diffusion model can be learned from traces of cascading failures, enabling accurate modeling and prediction of failure propagation. This approach facilitates actionable information through well-understood and efficient graph analysis methods and graph diffusion simulations. Furthermore, by leveraging the hyperparametric model, we can predict diffusion and mitigate the risks of cascading failures even in unseen grid configurations, whereas existing methods falter due to a lack of training data. Extensive experiments based on a benchmark power grid and simulations therein show that our approach effectively captures the failure diffusion phenomena and guides decisions to strengthen the grid, reducing the risk of large-scale cascading failures. Additionally, we characterize our model's sample complexity, improving upon the existing bound., Comment: KDD 2024

The two-phase horizontally periodic quasistationary Stokes flow in $\mathbb{R}^2$, describing the motion of two immiscible fluids with equal viscosities that are separated by a sharp interface, which is parameterized as the graph of a function $f=f(t)$, is considered in the general case when both gravity and surface tension effects are included. Using potential theory, the moving boundary problem is formulated as a fully nonlinear and nonlocal parabolic problem for the function $f$. Based on abstract parabolic theory, it is proven that the problem is well-posed in all subcritical spaces $\mathrm{H}^r(\mathbb{S})$, $r\in(3/2,2)$. Moreover, the stability properties of the flat equilibria are analyzed in dependence on the physical properties of the fluids., Comment: 39 pages

Using first-principles calculations, we explore the magnetoelectric properties of the room-temperature multiferroic crystal BiCoO$_3$. We use both applied magnetic field and finite-difference techniques to show that BiCoO$_3$ is anti-magnetoelectric at the linear level. The calculation of the dynamical effective charges reveals that the total magnetoelectric response is zero due to the compensating non-zero magnetoelectric response of each magnetic sublattice. This calculation also highlights that the the orbital contribution to the response is remarkably larger than the spin one and that each sublattice has a rather large total magnetoelectric response of 85 ps/m. Furthermore, we provide an intuitive recipe to visualize the dynamical magnetic effective charge, allowing to examine its multipolar nature which we confirm by means of ab initio calculations. Given the large value of the local response, we investigate the ferromagnetic phase as well, which gives a giant magnetoelectric response of about 1000 ps/m and coming mainly from the spin contribution this time. Finally, we discuss the possible reasons for such a large magnetoelectric response in BiCoO3 and propose possible strategies to unveil this potentially large response.

Recent research identified a temporary performance drop on previously learned tasks when transitioning to a new one. This drop is called the stability gap and has great consequences for continual learning: it complicates the direct employment of continually learning since the worse-case performance at task-boundaries is dramatic, it limits its potential as an energy-efficient training paradigm, and finally, the stability drop could result in a reduced final performance of the algorithm. In this paper, we show that the stability gap also occurs when applying joint incremental training of homogeneous tasks. In this scenario, the learner continues training on the same data distribution and has access to all data from previous tasks. In addition, we show that in this scenario, there exists a low-loss linear path to the next minima, but that SGD optimization does not choose this path. We perform further analysis including a finer batch-wise analysis which could provide insights towards potential solution directions., Comment: Accepted at CVPR 2024 Workshop on Continual Learning in Computer Vision (CLVision)

Deep learning has emerged as a powerful method for extracting valuable information from large volumes of data. However, when new training data arrives continuously (i.e., is not fully available from the beginning), incremental training suffers from catastrophic forgetting (i.e., new patterns are reinforced at the expense of previously acquired knowledge). Training from scratch each time new training data becomes available would result in extremely long training times and massive data accumulation. Rehearsal-based continual learning has shown promise for addressing the catastrophic forgetting challenge, but research to date has not addressed performance and scalability. To fill this gap, we propose an approach based on a distributed rehearsal buffer that efficiently complements data-parallel training on multiple GPUs, allowing us to achieve short runtime and scalability while retaining high accuracy. It leverages a set of buffers (local to each GPU) and uses several asynchronous techniques for updating these local buffers in an embarrassingly parallel fashion, all while handling the communication overheads necessary to augment input mini-batches (groups of training samples fed to the model) using unbiased, global sampling. In this paper we explore the benefits of this approach for classification models. We run extensive experiments on up to 128 GPUs of the ThetaGPU supercomputer to compare our approach with baselines representative of training-from-scratch (the upper bound in terms of accuracy) and incremental training (the lower bound). Results show that rehearsal-based continual learning achieves a top-5 classification accuracy close to the upper bound, while simultaneously exhibiting a runtime close to the lower bound., Comment: 2024 IEEE 24th International Symposium on Cluster, Cloud and Internet Computing (CCGrid), May 2024, Philadelphia (PA), United States

The paper presents the kinematic modelling for the coupled motion of a 6-DOF surgical parallel robot PARA-SILSROB which guides a mobile platform carrying the surgical instruments, and the actuators of the sub-modules which hold these tools. To increase the surgical procedure safety, a closed form solution for the kinematic model is derived and then, the forward and inverse kinematic models for the mobile orientation platform are obtained. The kinematic models are used in numerical simulations for the reorientation of the endoscopic camera, which imposes an automated compensatory motion from the active instruments' mod-ules.

Creativity has always been considered a major differentiator to separate the good from the great, and we believe the importance of creativity for software development will only increase as GenAI becomes embedded in developer tool-chains and working practices. This paper uses the McLuhan tetrad alongside scenarios of how GenAI may disrupt software development more broadly, to identify potential impacts GenAI may have on creativity within software development. The impacts are discussed along with a future research agenda comprising six connected themes that consider how individual capabilities, team capabilities, the product, unintended consequences, society, and human aspects can be affected.

Commercially impactful quantum algorithms such as quantum chemistry and Shor's algorithm require a number of qubits and gates far beyond the capacity of any existing quantum processor. Distributed architectures, which scale horizontally by networking modules, provide a route to commercial utility and will eventually surpass the capability of any single quantum computing module. Such processors consume remote entanglement distributed between modules to realize distributed quantum logic. Networked quantum computers will therefore require the capability to rapidly distribute high fidelity entanglement between modules. Here we present preliminary demonstrations of some key distributed quantum computing protocols on silicon T centres in isotopically-enriched silicon. We demonstrate the distribution of entanglement between modules and consume it to apply a teleported gate sequence, establishing a proof-of-concept for T centres as a distributed quantum computing and networking platform., Comment: 14 pages, 13 figures

Although supermassive black holes (SMBHs) reside in the heart of virtually every massive galaxy, it remains debated whether dwarf galaxies also commonly host SMBHs. Because low-mass galaxies may retain a memory of the assembly history of their black holes, probing the black hole occupation fraction of local dwarf galaxies might offer insights into the growth and seeding mechanisms of the first black holes. In this work, we exploit the Western half of the eROSITA all-sky survey (covering $20,000~\rm{deg^2}$) and compile a catalog of accreting SMBHs in local ($D<200$~Mpc) dwarf galaxies. After cleaning our sample from cosmic X-ray background sources, X-ray binaries, and ultraluminous X-ray sources, we identify 74 AGN-dwarf galaxy pairs. Using this large and uniform sample, we derive a luminosity function of dwarf galaxy AGN, fitting it with a power law function and obtaining ${\rm d}N/{\rm d}L_{\rm X} = (15.9\pm2.2)\times L_{\rm X}^{-1.63\pm0.05}$. Measuring the offset between the centroid of dwarf galaxies and the X-ray sources, we find that about $50\%$ of the AGN are likely off-nuclear, in agreement with theoretical predictions. We also compare the black hole-to-stellar mass relation of the AGN in our sample with the local and high-redshift relations, finding that our sources better adhere to the former. This suggests that local AGN across different mass scales underwent a similar growth history. Finally, we compare our sources with semi-analytical models: while our sample is too shallow to distinguish between different seeding models, it favors a growth mechanism linked to the star-formation rate of the host galaxy., Comment: 17 pages, 12 figures, 1 table. Submitted to ApJ. Comments are welcome

Boron nitride exhibits diverse crystal structures, predominantly a layered arrangement with strong intraplanar covalent bonds and weak interplanar van der Waals bonds. While commonly referred to as hexagonal BN (hBN), the sp$^2$-bonded BN atomic planes can also arrange in other configurations like Bernal (bBN) or rhombohedral (rBN) stacking orders. Variations in the orientation and translation of successive atomic layers lead to changes in crystal symmetry, potentially resulting in piezoelectric, pyroelectric or ferroelectric effects. However, distinguishing between different polytypes using conventional methods like X-ray diffraction or Raman spectroscopy presents a significant challenge. In this work, we demonstrate that the optical response of the 4.1 eV defect can serve as an indicator of the polytype. To this end, we study BN samples grown by metalorganic vapor phase epitaxy (MOVPE), which contain different polytypes. The identification of the polytypes was achieved by X-ray diffraction and transmission electron microscopy. Photoluminescence and cathodoluminescence measurements with a high spatial resolution allowed for the deconvolution of the signal into two components from which we can extract a zero-phonon line (ZPL) at 4.096 eV (302.6 nm) for hBN and 4.143 eV (299.2 nm) for rBN. We performed calculations that enable us to identify the defect as a carbon dimer CBCN (C2) and show that the ZPL shift reflects differences in the crystal environment for different polytypes. Furthermore, we demonstrate that different polytypic composition ratios of hBN and rBN can be achieved by MOVPE, which could pave the way for future applications in large-area van der Waals heterostructures.

We introduce Poseidon, a foundation model for learning the solution operators of PDEs. It is based on a multiscale operator transformer, with time-conditioned layer norms that enable continuous-in-time evaluations. A novel training strategy leveraging the semi-group property of time-dependent PDEs to allow for significant scaling-up of the training data is also proposed. Poseidon is pretrained on a diverse, large scale dataset for the governing equations of fluid dynamics. It is then evaluated on a suite of 15 challenging downstream tasks that include a wide variety of PDE types and operators. We show that Poseidon exhibits excellent performance across the board by outperforming baselines significantly, both in terms of sample efficiency and accuracy. Poseidon also generalizes very well to new physics that is not seen during pretraining. Moreover, Poseidon scales with respect to model and data size, both for pretraining and for downstream tasks. Taken together, our results showcase the surprising ability of Poseidon to learn effective representations from a very small set of PDEs during pretraining in order to generalize well to unseen and unrelated PDEs downstream, demonstrating its potential as an effective, general purpose PDE foundation model. Finally, the Poseidon model as well as underlying pretraining and downstream datasets are open sourced, with code being available at https://github.com/camlab-ethz/poseidon and pretrained models and datasets at https://huggingface.co/camlab-ethz.

Gaussian Splatting has revolutionized the world of novel view synthesis by achieving high rendering performance in real-time. Recently, studies have focused on enriching these 3D representations with semantic information for downstream tasks. In this paper, we introduce RT-GS2, the first generalizable semantic segmentation method employing Gaussian Splatting. While existing Gaussian Splatting-based approaches rely on scene-specific training, RT-GS2 demonstrates the ability to generalize to unseen scenes. Our method adopts a new approach by first extracting view-independent 3D Gaussian features in a self-supervised manner, followed by a novel View-Dependent / View-Independent (VDVI) feature fusion to enhance semantic consistency over different views. Extensive experimentation on three different datasets showcases RT-GS2's superiority over the state-of-the-art methods in semantic segmentation quality, exemplified by a 8.01% increase in mIoU on the Replica dataset. Moreover, our method achieves real-time performance of 27.03 FPS, marking an astonishing 901 times speedup compared to existing approaches. This work represents a significant advancement in the field by introducing, to the best of our knowledge, the first real-time generalizable semantic segmentation method for 3D Gaussian representations of radiance fields., Comment: Accepted paper at BMVC 2024

JavaScript packages are notoriously prone to bloat, a factor that significantly impacts the performance and maintainability of web applications. While web bundlers and tree-shaking can mitigate this issue in client-side applications at the function level, they cannot effectively detect and remove bloat in server-side applications. In this paper, we conduct an empirical study to investigate the bloated dependencies that are entirely unused within server-side applications. Our study focuses on applications built with the widely used and highly dynamic CommonJS module system. We propose a trace-based dynamic analysis that monitors file access, to determine which dependencies are not accessed during runtime. To conduct our study, we curate an original dataset of 92 CommonJS packages with a median test coverage of 96.9% and a total of 50,661 dependencies. Our dynamic analysis identifies and successfully removes 50.7% of these dependencies while maintaining the correct build of all packages. Furthermore, we find that 14.9% of directly used dependencies and 51.3% of indirect dependencies are bloated. A key insight is that focusing on removing only the direct bloated dependencies by cleaning the package.json file, also removes a significant share of unnecessary bloated indirect dependencies. Compared to the state-of-the-art dynamic debloating technique, our analysis based on file accesses has fewer false positives, and demonstrates higher accuracy in detecting bloated dependencies. Our findings suggest that native support for dependency debloating in package managers could significantly alleviate the burden of maintaining dependencies., Comment: Manuscript submitted to Empirical Software Engineering (EMSE)

This work studies embedding of arbitrary VC classes in well-behaved VC classes, focusing particularly on extremal classes. Our main result expresses an impossibility: such embeddings necessarily require a significant increase in dimension. In particular, we prove that for every $d$ there is a class with VC dimension $d$ that cannot be embedded in any extremal class of VC dimension smaller than exponential in $d$. In addition to its independent interest, this result has an important implication in learning theory, as it reveals a fundamental limitation of one of the most extensively studied approaches to tackling the long-standing sample compression conjecture. Concretely, the approach proposed by Floyd and Warmuth entails embedding any given VC class into an extremal class of a comparable dimension, and then applying an optimal sample compression scheme for extremal classes. However, our results imply that this strategy would in some cases result in a sample compression scheme at least exponentially larger than what is predicted by the sample compression conjecture. The above implications follow from a general result we prove: any extremal class with VC dimension $d$ has dual VC dimension at most $2d+1$. This bound is exponentially smaller than the classical bound $2^{d+1}-1$ of Assouad, which applies to general concept classes (and is known to be unimprovable for some classes). We in fact prove a stronger result, establishing that $2d+1$ upper bounds the dual Radon number of extremal classes. This theorem represents an abstraction of the classical Radon theorem for convex sets, extending its applicability to a wider combinatorial framework, without relying on the specifics of Euclidean convexity. The proof utilizes the topological method and is primarily based on variants of the Topological Radon Theorem.

Desirable random graph models (RGMs) should (i) be tractable so that we can compute and control graph statistics, and (ii) generate realistic structures such as high clustering (i.e., high subgraph densities). A popular category of RGMs (e.g., Erdos-Renyi and stochastic Kronecker) outputs edge probabilities, and we need to realize (i.e., sample from) the edge probabilities to generate graphs. Typically, each edge (in)existence is assumed to be determined independently. However, with edge independency, RGMs theoretically cannot produce high subgraph densities unless they "replicate" input graphs. In this work, we explore realization beyond edge independence that can produce more realistic structures while ensuring high tractability. Specifically, we propose edge-dependent realization schemes called binding and derive closed-form tractability results on subgraph (e.g., triangle) densities in graphs generated with binding. We propose algorithms for graph generation with binding and parameter fitting of binding. We empirically validate that binding exhibits high tractability and generates realistic graphs with high clustering, significantly improving upon existing RGMs assuming edge independency.

It has recently been demonstrated that pretraining backbones in a self-supervised manner generally provides better fine-tuned polyp segmentation performance, and that models with ViT-B backbones typically perform better than models with ResNet50 backbones. In this paper, we extend this recent work to consider generalisability. I.e., we assess the performance of models on a different dataset to that used for fine-tuning, accounting for variation in network architecture and pretraining pipeline (algorithm and dataset). This reveals how well models with different pretrained backbones generalise to data of a somewhat different distribution to the training data, which will likely arise in deployment due to different cameras and demographics of patients, amongst other factors. We observe that the previous findings, regarding pretraining pipelines for polyp segmentation, hold true when considering generalisability. However, our results imply that models with ResNet50 backbones typically generalise better, despite being outperformed by models with ViT-B backbones in evaluation on the test set from the same dataset used for fine-tuning.

Existing approaches for device placement ignore the topological features of computation graphs and rely mostly on heuristic methods for graph partitioning. At the same time, they either follow a grouper-placer or an encoder-placer architecture, which requires understanding the interaction structure between code operations. To bridge the gap between encoder-placer and grouper-placer techniques, we propose a novel framework for the task of device placement, relying on smaller computation graphs extracted from the OpenVINO toolkit using reinforcement learning. The framework consists of five steps, including graph coarsening, node representation learning and policy optimization. It facilitates end-to-end training and takes into consideration the directed and acyclic nature of the computation graphs. We also propose a model variant, inspired by graph parsing networks and complex network analysis, enabling graph representation learning and personalized graph partitioning jointly, using an unspecified number of groups. To train the entire framework, we utilize reinforcement learning techniques by employing the execution time of the suggested device placements to formulate the reward. We demonstrate the flexibility and effectiveness of our approach through multiple experiments with three benchmark models, namely Inception-V3, ResNet, and BERT. The robustness of the proposed framework is also highlighted through an ablation study. The suggested placements improve the inference speed for the benchmark models by up to $58.2\%$ over CPU execution and by up to $60.24\%$ compared to other commonly used baselines.

We study the power of (competitive) algorithms with predictions in a multiagent setting. For this we introduce a multiagent version of the ski-rental problem. In this problem agents can collaborate by pooling resources to get a group license for some asset. If the license price is not met agents have to rent the asset individually for the day at a unit price. Otherwise the license becomes available forever to everyone at no extra cost. Our main contribution is a best-response analysis of a single-agent competitive algorithm that assumes perfect knowledge of other agents' actions (but no knowledge of its own renting time). We then analyze the setting when agents have a predictor for their own active time, yielding a tradeoff between robustness and consistency. We investigate the effect of using such a predictor in an equilibrium, as well as the new equilibria formed in this way.

This paper presents a design and implementation of a high-power Gallium Nitride (GaN)-based multilevel Hbridge inverter to excite wireless charging coils for the wireless power transfer of electric vehicles (EVs). Compared to the traditional conductive charging, wireless charging technology offers a safer and more convenient way to charge EVs. Due to the increasing demand of fast charging, high-power inverters play a crucial role in exciting the wireless charging coils within a wireless power transfer system. This paper details the system specifications for the wireless charging of EVs, providing theoretical analysis and a control strategy for the modular design of a 75-kW 3-level and 4-level Hbridge inverter. The goal is to deliver a low-distortion excitation voltage to the wireless charging coils. LTspice simulation results, including output voltage, Fast Fourier Transform (FFT) analysis for both 3-level and 4-level H-bridge inverters, are presented to validate the control strategy and demonstrate the elimination of output harmonic components in the modular design. A GaNbased inverter prototype was employed to deliver a 85-kHz power to the wireless charging pads of the wireless power transfer system. Experimental results at two different voltage and power levels, 100V-215W and 150V-489W, validate the successful performance of the GaN inverter in the wireless charging system.

In this paper, we provide a representation of a certain class of C*-valued positive sesquilinear and linear maps on non-unital quasi *-algebras. Also, we illustrate our results on the concrete examples of non-unital Banach quasi *-algebras, such as the standard Hilbert module over a commutative C*-algebra, Schatten p-ideals, and noncommutative L2 spaces induced by a semifinite, nonfinite trace. As a consequence of our results, we obtain a representation of all bounded positive linear C*-valued maps on non-unital C*-algebras. We also deduce some norm inequalities for these maps. Finally, we consider a noncommutative L2 space equipped with the topology generated by a positive sesquilinear form and we construct a topologically transitive operator on this space., Comment: This version of the article has been accepted for publication, after peer review (when applicable) but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: http://dx.doi.org/10.1007/s11117-024-01079-8

Consider a graph on the non-singular matrices over a finite field, in which two distinct non-singular matrices are joined by an edge whenever their sum is singular. We prove an upper bound for the independence number of this graph. As a consequence, we obtain a lower bound for its chromatic number that significantly improves a previous result of Tomon.

We obtain asymptotic estimates for the $\ell^p$-operator norm of spherical averaging operators associated to certain geometric group actions. The motivating example is the case of Gromov hyperbolic groups, for which we obtain asymptotically sharp estimates. We deduce asymptotic lower bounds for the combinatorial expansion of spheres., Comment: 20 pages

Popular regularizers with non-differentiable penalties, such as Lasso, Elastic Net, Generalized Lasso, or SLOPE, reduce the dimension of the parameter space by inducing sparsity or clustering in the estimators' coordinates. In this paper, we focus on linear regression and explore the asymptotic distributions of the resulting low-dimensional patterns when the number of regressors $p$ is fixed, the number of observations $n$ goes to infinity, and the penalty function increases at the rate of $\sqrt{n}$. While the asymptotic distribution of the rescaled estimation error can be derived by relatively standard arguments, the convergence of the pattern does not simply follow from the convergence in distribution, and requires a careful and separate treatment. For this purpose, we use the Hausdorff distance as a suitable mode of convergence for subdifferentials, resulting in the desired pattern convergence. Furthermore, we derive the exact limiting probability of recovering the true model pattern. This probability goes to 1 if and only if the penalty scaling constant diverges to infinity and the regularizer-specific asymptotic irrepresentability condition is satisfied. We then propose simple two-step procedures that asymptotically recover the model patterns, irrespective whether the irrepresentability condition holds. Interestingly, our theory shows that Fused Lasso cannot reliably recover its own clustering pattern, even for independent regressors. It also demonstrates how this problem can be resolved by ``concavifying'' the Fused Lasso penalty coefficients. Additionally, sampling from the asymptotic error distribution facilitates comparisons between different regularizers. We provide short simulation studies showcasing an illustrative comparison between the asymptotic properties of Lasso, Fused Lasso, and SLOPE., Comment: 30 pages, 5 figures, for associated code, see https://github.com/IvanHejny/pyslope.git

Static soliton bound states in nonlinear systems are investigated analytically and numerically in the framework of the parametrically driven, damped nonlinear Schr\"odinger equation. We find that the ordinary differential equations, which determine bound soliton solutions, can be transformed into the form resembling the Schr\"odinger-like equations for eigenfunctions with the fixed eigenvalues. We assume that a nonlinear part of the equations is close to the reflectionless potential well occurring in the scattering problem, associated with the integrable equations. We show that symmetric two-hump soliton solution is quite well described analytically by the three-soliton formula with the fixed soliton parameters, depending on the strength of parametric pumping and the dissipation constant., Comment: 20 pages, 4 figures

The numerical solution of problems in nonlinear magnetostatics is typically based on a variational formulation in terms of magnetic potentials, the discretization by finite elements, and iterative solvers like the Newton method. The vector potential approach aims at minimizing a certain energy functional and, in three dimensions, requires the use of edge elements and appropriate gauging conditions. The scalar potential approach, on the other hand, seeks to maximize the negative coenergy and can be realized by standard Lagrange finite elements, thus reducing the number of degrees of freedom and simplifying the implementation. The number of Newton iterations required to solve the governing nonlinear system, however, has been observed to be usually higher than for the vector potential formulation. In this paper, we propose a method that combines the advantages of both approaches, i.e., it requires as few Newton iterations as the vector potential formulation while involving the magnetic scalar potential as the primary unknown. We discuss the variational background of the method, its well-posedness, and its efficient implementation. Numerical examples are presented for illustration of the accuracy and the gain in efficiency compared to other approaches.

The discovery of Little Red Dots (LRDs) -- a population of compact, high-redshift, dust-reddened galaxies -- is one of the most surprising results from JWST. However, the nature of LRDs is still debated: does the near-infrared emission originate from accreting supermassive black holes (SMBHs), or intense star formation? In this work, we utilize ultra-deep Chandra observations and study LRDs residing behind the lensing galaxy cluster, Abell~2744. We probe the X-ray emission from individual galaxies but find that they remain undetected and provide SMBH mass upper limits of $\lesssim(1.5-16)\times10^{6}~\rm{M_{\odot}}$ assuming Eddington limited accretion. To increase the signal-to-noise ratios, we conduct a stacking analysis of the full sample with a total lensed exposure time of $\approx87$~Ms. We also bin the galaxies based on their stellar mass, lensing magnification, and detected broad-line H$\alpha$ emission. For the LRDs exhibiting broad-line H$\alpha$ emission, there is a hint of a stacked signal ($\sim2.6\sigma$), corresponding to a SMBH mass of $\sim3.2\times10^{6}~\rm{M_{\odot}}$. Assuming unobscured, Eddington-limited accretion, this BH mass is at least 1.5 orders of magnitude lower than that inferred from virial mass estimates using JWST spectra. Given galaxy-dominated stellar mass estimates, our results imply that LRDs do not host over-massive SMBHs and/or accrete at a few percent of their Eddington limit. However, alternative stellar mass estimates may still support that LRDs host over-massive BHs. The significant discrepancy between the JWST and Chandra data hints that the scaling relations used to infer the SMBH mass from the H$\alpha$ line and virial relations may not be applicable for high-redshift LRDs., Comment: Accepted by ApJL; DOI: 10.3847/2041-8213/ad5669

The paper explores the use of dynamic adaptation gain/step size (DAG) for improving the adaptation transient performance of variable step-size LMS (VS-LMS) adaptation algorithms. A generic form for the implementation of the DAG within the VS-LMS algorithms is provided. The properties of the VS-LMS algorithms using dynamic adaptation gain are discussed in detail. Stability issues in deterministic environment and convergence properties in stochastic environment are examined. A transient performance analysis is proposed. Criteria for the selection of the coefficients of the DAG filter are provided.The potential of the VS-LMS adaptation algorithms using a DAG is then illustrated by simulation results (adaptive line enhancer, filter identification) and experimental results obtained on a relevant adaptive active noise attenuation system., Comment: arXiv admin note: substantial text overlap with arXiv:2403.13381

The paper presents the methodology used for accuracy and repeatability measurements of the experimental model of a parallel robot developed for surgical applications. The experimental setup uses a motion tracking system (for accuracy) and a high precision measuring arm for position (for repeatability). The accuracy was obtained by comparing the trajectory data from the experimental measurement with a baseline trajectory defined with the kinematic models of the parallel robotic system. The repeatability was experi-mentally determined by moving (repeatedly) the robot platform in predefined points.

It is shown that semilinear parabolic evolution equations $u'=A+f(t,u)$ featuring H\"older continuous nonlinearities $ f=f(t,u)$ with at most linear growth possess global strong solutions for a general class of initial data. The abstract results are applied to a recent model describing front propagation in bushfires and in the context of a reaction-diffusion system., Comment: 15 pages

Integrating and processing information from various sources or modalities are critical for obtaining a comprehensive and accurate perception of the real world in autonomous systems and cyber-physical systems. Drawing inspiration from neuroscience, we develop the Information-Theoretic Hierarchical Perception (ITHP) model, which utilizes the concept of information bottleneck. Different from most traditional fusion models that incorporate all modalities identically in neural networks, our model designates a prime modality and regards the remaining modalities as detectors in the information pathway, serving to distill the flow of information. Our proposed perception model focuses on constructing an effective and compact information flow by achieving a balance between the minimization of mutual information between the latent state and the input modal state, and the maximization of mutual information between the latent states and the remaining modal states. This approach leads to compact latent state representations that retain relevant information while minimizing redundancy, thereby substantially enhancing the performance of multimodal representation learning. Experimental evaluations on the MUStARD, CMU-MOSI, and CMU-MOSEI datasets demonstrate that our model consistently distills crucial information in multimodal learning scenarios, outperforming state-of-the-art benchmarks. Remarkably, on the CMU-MOSI dataset, ITHP surpasses human-level performance in the multimodal sentiment binary classification task across all evaluation metrics (i.e., Binary Accuracy, F1 Score, Mean Absolute Error, and Pearson Correlation)., Comment: Accepted by ICLR 2024. Camera Ready Version

This paper collects polynomial Diophantine equations that are amazingly simple to write down but are apparently difficult to solve.

Audio-visual zero-shot learning methods commonly build on features extracted from pre-trained models, e.g. video or audio classification models. However, existing benchmarks predate the popularization of large multi-modal models, such as CLIP and CLAP. In this work, we explore such large pre-trained models to obtain features, i.e. CLIP for visual features, and CLAP for audio features. Furthermore, the CLIP and CLAP text encoders provide class label embeddings which are combined to boost the performance of the system. We propose a simple yet effective model that only relies on feed-forward neural networks, exploiting the strong generalization capabilities of the new audio, visual and textual features. Our framework achieves state-of-the-art performance on VGGSound-GZSL, UCF-GZSL, and ActivityNet-GZSL with our new features. Code and data available at: https://github.com/dkurzend/ClipClap-GZSL., Comment: CVPRw 2024 (L3D-IVU)