arXiv subjects

Kai Wang

Publications and source records attributed to Kai Wang.

At least 415 records · Page 23Linked to original sources

arXiv2024

ERCache: An Efficient and Reliable Caching Framework for Large-Scale User Representations in Meta's Ads System

The increasing complexity of deep learning models used for calculating user representations presents significant challenges, particularly with limited computational resources and strict service-level agreements (SLAs). Previous research efforts have focused on optimizing model inference but have overlooked a critical question: is it necessary to perform user model inference for every ad request in large-scale social networks? To address this question and these challenges, we first analyze user access patterns at Meta and find that most user model inferences occur within a short timeframe. T his observation reveals a triangular relationship among model complexity, embedding freshness, and service SLAs. Building on this insight, we designed, implemented, and evaluated ERCache, an efficient and robust caching framework for large-scale user representations in ads recommendation systems on social networks. ERCache categorizes cache into direct and failover types and applies customized settings and eviction policies for each model, effectively balancing model complexity, embedding freshness, and service SLAs, even considering the staleness introduced by caching. ERCache has been deployed at Meta for over six months, supporting more than 30 ranking models while efficiently conserving computational resources and complying with service SLA requirements.

cs.IR↗

arXiv2024

The bounded ideal monad on the category of quasi-metric spaces and its algebras

The notion of bounded ideals is introduced for quasi-metric spaces. Such ideals give rise to a monad, the bounded ideal monad, on the category of quasi-metric spaces and non-expansive maps. Algebras of this monad are metric version of local dcpos of Mislove. It is shown that an algebra of the bounded ideal monad is a standard quasi-metric space of which the formal balls form a local dcpo; and that a continuous algebra is a standard quasi-metric space of which the formal balls form a local domain.

math.CT↗

arXiv2024

Two Trades is not Baffled: Condensing Graph via Crafting Rational Gradient Matching

Training on large-scale graphs has achieved remarkable results in graph representation learning, but its cost and storage have raised growing concerns. As one of the most promising directions, graph condensation methods address these issues by employing gradient matching, aiming to condense the full graph into a more concise yet information-rich synthetic set. Though encouraging, these strategies primarily emphasize matching directions of the gradients, which leads to deviations in the training trajectories. Such deviations are further magnified by the differences between the condensation and evaluation phases, culminating in accumulated errors, which detrimentally affect the performance of the condensed graphs. In light of this, we propose a novel graph condensation method named \textbf{C}raf\textbf{T}ing \textbf{R}ationa\textbf{L} trajectory (\textbf{CTRL}), which offers an optimized starting point closer to the original dataset's feature distribution and a more refined strategy for gradient matching. Theoretically, CTRL can effectively neutralize the impact of accumulated errors on the performance of condensed graphs. We provide extensive experiments on various graph datasets and downstream tasks to support the effectiveness of CTRL. Code is released at https://github.com/NUS-HPC-AI-Lab/CTRL.

cs.LG↗

arXiv2024

Electromagnetic Flares Associated with Gravitational Waves from Binary Black Hole Mergers in AGN Accretion Disks

The gravitational wave (GW) event GW190521, likely originating from a binary black hole (BBH) merger within an active galactic nucleus (AGN) disk, is associated with the optical flare ZTF19abanrhr. The remnant BHs from BBH mergers can launch the jet and outflow and then interact with the disk medium, which can be responsible for the associated electromagnetic radiations. In this \textit{letter}, we examine the shock breakout and subsequent cooling emissions from four potential components: the outflow, jet head, jet cocoon, and disk cocoon, all driven by the remnant BH within the AGN disk. Using dynamic models and observational constraints, for GW190521, we identify the parameter space for each component and conclude that either the outflow or the disk cocoon could produce the observed electromagnetic signal, with the disk cocoon requiring more extreme parameters. We present best-fit light curves and spectral energy distributions (SEDs) for both components, showing peak emissions in the UV band for the outflow and spanning optical to UV for the disk cocoon.

astro-ph.HE↗

arXiv2024

Robustifying Model Predictive Control of Uncertain Linear Systems with Chance Constraints

This paper proposes a model predictive controller for discrete-time linear systems with additive, possibly unbounded, stochastic disturbances and subject to chance constraints. By computing a polytopic probabilistic positively invariant set for constraint tightening with the help of the computation of the minimal robust positively invariant set, the chance constraints are guaranteed, assuming only the mean and covariance of the disturbance distribution are given. The resulting online optimization problem is a standard strictly quadratic programming, just like in conventional model predictive control with recursive feasibility and stability guarantees and is simple to implement. A numerical example is provided to illustrate the proposed method.

math.OC↗

arXiv2024

CHiSafetyBench: A Chinese Hierarchical Safety Benchmark for Large Language Models

With the profound development of large language models(LLMs), their safety concerns have garnered increasing attention. However, there is a scarcity of Chinese safety benchmarks for LLMs, and the existing safety taxonomies are inadequate, lacking comprehensive safety detection capabilities in authentic Chinese scenarios. In this work, we introduce CHiSafetyBench, a dedicated safety benchmark for evaluating LLMs' capabilities in identifying risky content and refusing answering risky questions in Chinese contexts. CHiSafetyBench incorporates a dataset that covers a hierarchical Chinese safety taxonomy consisting of 5 risk areas and 31 categories. This dataset comprises two types of tasks: multiple-choice questions and question-answering, evaluating LLMs from the perspectives of risk content identification and the ability to refuse answering risky questions respectively. Utilizing this benchmark, we validate the feasibility of automatic evaluation as a substitute for human evaluation and conduct comprehensive automatic safety assessments on mainstream Chinese LLMs. Our experiments reveal the varying performance of different models across various safety domains, indicating that all models possess considerable potential for improvement in Chinese safety capabilities. Our dataset is publicly available at https://github.com/UnicomAI/UnicomBenchmark/tree/main/CHiSafetyBench.

cs.CL↗

arXiv2024

NGEL-SLAM: Neural Implicit Representation-based Global Consistent Low-Latency SLAM System

Neural implicit representations have emerged as a promising solution for providing dense geometry in Simultaneous Localization and Mapping (SLAM). However, existing methods in this direction fall short in terms of global consistency and low latency. This paper presents NGEL-SLAM to tackle the above challenges. To ensure global consistency, our system leverages a traditional feature-based tracking module that incorporates loop closure. Additionally, we maintain a global consistent map by representing the scene using multiple neural implicit fields, enabling quick adjustment to the loop closure. Moreover, our system allows for fast convergence through the use of octree-based implicit representations. The combination of rapid response to loop closure and fast convergence makes our system a truly low-latency system that achieves global consistency. Our system enables rendering high-fidelity RGB-D images, along with extracting dense and complete surfaces. Experiments on both synthetic and real-world datasets suggest that our system achieves state-of-the-art tracking and mapping accuracy while maintaining low latency.

cs.RO↗

arXiv2024

Exceptional swallowtail degeneracies in driven-dissipative quadrature squeezing

We show that swallowtail catastrophe consisting of various-order non-Hermitian and Hermitian degeneracies naturally exists in the dynamics of two-mode driven-dissipative quadrature squeezing systems that break pseudo-Hermiciticy by judiciously engineered losses. We reveal that the swallowtail degeneracy structure enables nontrivial braiding of complex eigenvalues by looping around an exceptional line in the parameter space. Our findings provide a comprehensive understanding of the degeneracy geometry in two-mode driven-dissipative bosonic quadratic systems, opening new pathways toward topologically nontrivial control of Gaussian states.

quant-ph↗

arXiv2024

Femtosecond pulse amplification on a chip

Femtosecond laser pulses enable the synthesis of light across the electromagnetic spectrum and provide access to ultrafast phenomena in physics, biology, and chemistry. Chip-integration of femtosecond technology could revolutionize applications such as point-of-care diagnostics, bio-medical imaging, portable chemical sensing, or autonomous navigation. However, current chip-integrated pulse sources lack the required peak power and on-chip amplification of femtosecond pulses has been an unresolved challenge. Here, addressing this challenge, we report >50-fold amplification of 1 GHz-repetition-rate chirped femtosecond pulses in a CMOS-compatible photonic chip to 800 W peak power with 116 fs pulse duration. This power level is 2-3 orders of magnitude higher compared to those in previously demonstrated on-chip pulse sources and can provide the power needed to address key applications. To achieve this, detrimental nonlinear effects are mitigated through all-normal dispersion, large mode-area and rare-earth-doped gain waveguides. These results offer a pathway to chip-integrated femtosecond technology with peak power-levels characteristic of table-top sources.

physics.optics↗

arXiv2024

GRIF-DM: Generation of Rich Impression Fonts using Diffusion Models

Fonts are integral to creative endeavors, design processes, and artistic productions. The appropriate selection of a font can significantly enhance artwork and endow advertisements with a higher level of expressivity. Despite the availability of numerous diverse font designs online, traditional retrieval-based methods for font selection are increasingly being supplanted by generation-based approaches. These newer methods offer enhanced flexibility, catering to specific user preferences and capturing unique stylistic impressions. However, current impression font techniques based on Generative Adversarial Networks (GANs) necessitate the utilization of multiple auxiliary losses to provide guidance during generation. Furthermore, these methods commonly employ weighted summation for the fusion of impression-related keywords. This leads to generic vectors with the addition of more impression keywords, ultimately lacking in detail generation capacity. In this paper, we introduce a diffusion-based method, termed \ourmethod, to generate fonts that vividly embody specific impressions, utilizing an input consisting of a single letter and a set of descriptive impression keywords. The core innovation of \ourmethod lies in the development of dual cross-attention modules, which process the characteristics of the letters and impression keywords independently but synergistically, ensuring effective integration of both types of information. Our experimental results, conducted on the MyFonts dataset, affirm that this method is capable of producing realistic, vibrant, and high-fidelity fonts that are closely aligned with user specifications. This confirms the potential of our approach to revolutionize font generation by accommodating a broad spectrum of user-driven design requirements. Our code is publicly available at \url{https://github.com/leitro/GRIF-DM}.

cs.CV↗

arXiv2024

Black holes regulate cool gas accretion in massive galaxies

The nucleus of almost all massive galaxies contains a supermassive black hole (BH). The feedback from the accretion of these BHs is often considered to have crucial roles in establishing the quiescence of massive galaxies, although some recent studies show that even galaxies hosting the most active BHs do not exhibit a reduction in their molecular gas reservoirs or star formation rates. Therefore, the influence of BHs on galaxy star formation remains highly debated and lacks direct evidence. Here, based on a large sample of nearby galaxies with measurements of masses of both BHs and atomic hydrogen (HI), the main component of the interstellar medium, we show that the HI gas mass to stellar masses ratio ($μ_{\rm HI} = M_{\rm HI}/M_{\star}$) is more strongly correlated with BH masses ($M_{\rm BH}$) than with any other galaxy parameters, including stellar mass, stellar mass surface density and bulge masses. Moreover, once the $μ_{\rm HI}-M_{\rm BH}$ correlation is considered, $μ_{\rm HI}$ loses dependence on other galactic parameters, demonstrating that $M_{\rm BH}$ serves as the primary driver of $μ_{\rm HI}$. These findings provide important evidence for how the accumulated energy from BH accretion regulates the cool gas content in galaxies, by ejecting interstellar medium gas and/or suppressing gas cooling from the circumgalactic medium.

astro-ph.GA↗

arXiv2024

More Than Positive and Negative: Communicating Fine Granularity in Medical Diagnosis

With the advance of deep learning, much progress has been made in building powerful artificial intelligence (AI) systems for automatic Chest X-ray (CXR) analysis. Most existing AI models are trained to be a binary classifier with the aim of distinguishing positive and negative cases. However, a large gap exists between the simple binary setting and complicated real-world medical scenarios. In this work, we reinvestigate the problem of automatic radiology diagnosis. We first observe that there is considerable diversity among cases within the positive class, which means simply classifying them as positive loses many important details. This motivates us to build AI models that can communicate fine-grained knowledge from medical images like human experts. To this end, we first propose a new benchmark on fine granularity learning from medical images. Specifically, we devise a division rule based on medical knowledge to divide positive cases into two subcategories, namely atypical positive and typical positive. Then, we propose a new metric termed AUC$^\text{FG}$ on the two subcategories for evaluation of the ability to separate them apart. With the proposed benchmark, we encourage the community to develop AI diagnosis systems that could better learn fine granularity from medical images. Last, we propose a simple risk modulation approach to this problem by only using coarse labels in training. Empirical results show that despite its simplicity, the proposed method achieves superior performance and thus serves as a strong baseline.

cs.CV↗

arXiv2024

CHORDination: Evaluating Visual Design Choices in Chord Diagrams for Network Data

Chord diagrams are widely used for visualizing data connectivity and flow between nodes in a network. They are effective for representing complex structures through an intuitive and visually appealing circular layout. While previous work has focused on improving aesthetics and interactivity, the influence of fundamental design elements on user perception and information retrieval remains under-explored. In this study, we explored the three primary components of chord diagram anatomy, namely the nodes, circular outline, and arc connections, in three sequential experiment phases. In phase one, we conducted a controlled experiment (N=90) to find the perceptually and information optimized node widths (narrow, medium, wide) and quantities (low, medium, high). This optimal set of node width and quantity sets the foundation for subsequent evaluations and were kept fixed for consistency. In phase two of the study, we conducted an expert design review for identifying the optimal radial tick marks and color gradients. Then in phase three, we evaluated the perceptual and information retrieval performance of the design choices in a controlled experiment (N=24) by comparing four chord diagram designs (baseline, radial tick marks, arc color gradients, both tick marks and color gradients). Results indicated that node width and quantity significantly affected users' information retrieval performance and subjective ratings, whereas the presence of tick marks predominantly influenced subjective experiences. Based on these findings, we discuss the design implications of these visual elements and offer guidance and recommendations for optimizing chord diagram designs in network visualization tasks.

cs.HC↗

arXiv2024

Traveling waves of NLS System arising in optical material without Galilean symmetry

We consider a system of NLS with cubic interactions arising in nonlinear optics without Galilean symmetry. The absence of Galilean symmetry can lead to many difficulties, such as global existence and blowup problems; see [Comm. Partial Differential Equations 46, 11 (2021), 2134-2170]. In this paper, we mainly focus on the influence of the absence of this symmetry on the traveling waves of the NLS system. Firstly, we obtain the existence of traveling solitary wave solutions that are non-radial and complex-valued. Secondly, using the asymptotic analysis method, when the frequency is sufficiently large, we establish the high frequency limit of the traveling solitary wave solution. Finally, for the mass critical case, we provide a novel condition for the existence of global solutions which is significantly different from the classical. In particular, this new condition breaks the traditional optimal assumption about initial data.

math.AP↗

arXiv2024

TS-SAM: Fine-Tuning Segment-Anything Model for Downstream Tasks

Adapter based fine-tuning has been studied for improving the performance of SAM on downstream tasks. However, there is still a significant performance gap between fine-tuned SAMs and domain-specific models. To reduce the gap, we propose Two-Stream SAM (TS-SAM). On the one hand, inspired by the side network in Parameter-Efficient Fine-Tuning (PEFT), we designed a lightweight Convolutional Side Adapter (CSA), which integrates the powerful features from SAM into side network training for comprehensive feature fusion. On the other hand, in line with the characteristics of segmentation tasks, we designed Multi-scale Refinement Module (MRM) and Feature Fusion Decoder (FFD) to keep both the detailed and semantic features. Extensive experiments on ten public datasets from three tasks demonstrate that TS-SAM not only significantly outperforms the recently proposed SAM-Adapter and SSOM, but achieves competitive performance with the SOTA domain-specific models. Our code is available at: https://github.com/maoyangou147/TS-SAM.

cs.CV↗

arXiv2024

Piculet: Specialized Models-Guided Hallucination Decrease for MultiModal Large Language Models

Multimodal Large Language Models (MLLMs) have made significant progress in bridging the gap between visual and language modalities. However, hallucinations in MLLMs, where the generated text does not align with image content, continue to be a major challenge. Existing methods for addressing hallucinations often rely on instruction-tuning, which requires retraining the model with specific data, which increases the cost of utilizing MLLMs further. In this paper, we introduce a novel training-free method, named Piculet, for enhancing the input representation of MLLMs. Piculet leverages multiple specialized models to extract descriptions of visual information from the input image and combine these descriptions with the original image and query as input to the MLLM. We evaluate our method both quantitively and qualitatively, and the results demonstrate that Piculet greatly decreases hallucinations of MLLMs. Our method can be easily extended to different MLLMs while being universal.

cs.AI↗

arXiv2024

Conditional LoRA Parameter Generation

Generative models have achieved remarkable success in image, video, and text domains. Inspired by this, researchers have explored utilizing generative models to generate neural network parameters. However, these efforts have been limited by the parameter size and the practicality of generating high-performance parameters. In this paper, we propose COND P-DIFF, a novel approach that demonstrates the feasibility of controllable high-performance parameter generation, particularly for LoRA (Low-Rank Adaptation) weights, during the fine-tuning process. Specifically, we employ an autoencoder to extract efficient latent representations for parameters. We then train a conditional latent diffusion model to synthesize high-performing model parameters from random noise based on specific task conditions. Experimental results in both computer vision and natural language processing domains consistently demonstrate that COND P-DIFF can generate high-performance parameters conditioned on the given task. Moreover, we observe that the parameter distribution generated by COND P-DIFF exhibits differences compared to the distribution obtained through normal optimization methods, indicating a certain level of generalization capability. Our work paves the way for further exploration of condition-driven parameter generation, offering a promising direction for task-specific adaptation of neural networks.

cs.AI↗

arXiv2024

Revisiting the fundamental metallicity relation with observation and simulation

The gas-phase metallicity of galaxies is regulated by multiple astrophysical processes, which makes it a crucial diagnostic of galaxy formation and evolution. Beyond the fundamental mass-metallicity relation, a debate about the secondary galaxy property to predict the metallicity of galaxies arises. Motivated by this, we systematically examine the relationship between gas-phase metallicity and other galaxy properties, i.e. star formation rate (SFR) and galaxy size, in addition to stellar mass in both observation and simulation. We utilize the data from the MaNGA (Mapping Nearby Galaxies at Apache Point Observatory) survey and the TNG50 simulations. We find that the combination of $M_*/R_{\rm e}^β$ with $β\sim 0.6-1$ is in much stronger correlation to the metallicity than stellar mass alone, regardless of whether the SFR is included or not, in both observation and simulation. This indicates that galaxy size plays a more important role in determining gas-phase metallicity of galaxies than SFR. In addition, the TNG simulation predicts that the SFR, although being a subdominant role, becomes increasingly important in high-$z$ universe. Finally, we speculate that SFR modulates metallicity on the temporal dimension, synchronized with time-varying gas inflows, and galaxy size regulates metallicity on the spatial dimension by affecting the gravitational potential and the mass loading factor.

astro-ph.GA↗