arXiv subjects

Yifan Zhu

Publications and source records attributed to Yifan Zhu.

At least 109 records · Page 6Linked to original sources

arXiv2025

Generalizability of Neural Networks Minimizing Empirical Risk Based on Expressive Ability

The primary objective of learning methods is generalization. Classic uniform generalization bounds, which rely on VC-dimension or Rademacher complexity, fail to explain the significant attribute that over-parameterized models in deep learning exhibit nice generalizability. On the other hand, algorithm-dependent generalization bounds, like stability bounds, often rely on strict assumptions. To establish generalizability under less stringent assumptions, this paper investigates the generalizability of neural networks that minimize or approximately minimize empirical risk. We establish a lower bound for population accuracy based on the expressiveness of these networks, which indicates that with an adequate large number of training samples and network sizes, these networks, including over-parameterized ones, can generalize effectively. Additionally, we provide a necessary condition for generalization, demonstrating that, for certain data distributions, the quantity of training data required to ensure generalization exceeds the network size needed to represent the corresponding data distribution. Finally, we provide theoretical insights into several phenomena in deep learning, including robust generalization, importance of over-parameterization, and effect of loss function on generalization.

cs.LG↗

arXiv2025

Provable Robust Overfitting Mitigation in Wasserstein Distributionally Robust Optimization

Wasserstein distributionally robust optimization (WDRO) optimizes against worst-case distributional shifts within a specified uncertainty set, leading to enhanced generalization on unseen adversarial examples, compared to standard adversarial training which focuses on pointwise adversarial perturbations. However, WDRO still suffers fundamentally from the robust overfitting problem, as it does not consider statistical error. We address this gap by proposing a novel robust optimization framework under a new uncertainty set for adversarial noise via Wasserstein distance and statistical error via Kullback-Leibler divergence, called the Statistically Robust WDRO. We establish a robust generalization bound for the new optimization framework, implying that out-of-distribution adversarial performance is at least as good as the statistically robust training loss with high probability. Furthermore, we derive conditions under which Stackelberg and Nash equilibria exist between the learner and the adversary, giving an optimal robust model in certain sense. Finally, through extensive experiments, we demonstrate that our method significantly mitigates robust overfitting and enhances robustness within the framework of WDRO.

cs.LG↗

arXiv2025

Accelerating Graph Indexing for ANNS on Modern CPUs

In high-dimensional vector spaces, Approximate Nearest Neighbor Search (ANNS) is a key component in database and artificial intelligence infrastructures. Graph-based methods, particularly HNSW, have emerged as leading solutions among various ANNS approaches, offering an impressive trade-off between search efficiency and accuracy. Many modern vector databases utilize graph indexes as their core algorithms, benefiting from various optimizations to enhance search performance. However, the high indexing time associated with graph algorithms poses a significant challenge, especially given the increasing volume of data, query processing complexity, and dynamic index maintenance demand. This has rendered indexing time a critical performance metric for users. In this paper, we comprehensively analyze the underlying causes of the low graph indexing efficiency on modern CPUs, identifying that distance computation dominates indexing time, primarily due to high memory access latency and suboptimal arithmetic operation efficiency. We demonstrate that distance comparisons during index construction can be effectively performed using compact vector codes at an appropriate compression error. Drawing from insights gained through integrating existing compact coding methods in the graph indexing process, we propose a novel compact coding strategy, named Flash, designed explicitly for graph indexing and optimized for modern CPU architectures. By minimizing random memory accesses and maximizing the utilization of SIMD (Single Instruction, Multiple Data) instructions, Flash significantly enhances cache hit rates and arithmetic operations. Extensive experiments conducted on eight real-world datasets, ranging from ten million to one billion vectors, exhibit that Flash achieves a speedup of 10.4$\times$ to 22.9$\times$ in index construction efficiency, while maintaining or improving search performance.

cs.DB↗

arXiv2025

A Stronger Mixture of Low-Rank Experts for Fine-Tuning Foundation Models

In order to streamline the fine-tuning of foundation models, Low-Rank Adapters (LoRAs) have been substantially adopted across various fields, including instruction tuning and domain adaptation. The underlying concept of LoRA involves decomposing a full-rank matrix into the product of two lower-rank matrices, which reduces storage consumption and accelerates the training process. Furthermore, to address the limited expressive capacity of LoRA, the Mixture-of-Expert (MoE) has been introduced for incorporating multiple LoRA adapters. The integration of LoRA experts leads to a visible improvement across several downstream scenes. However, the mixture of LoRAs (MoE-LoRA) still exhibits its low robustness during tuning and inferring. Inspired by the Riemannian Preconditioners which train LoRA as a sub-space projector, we propose a new training strategy for MoE-LoRA, to stabilize and boost its feature learning procedure by multi-space projections. Examinations on SGD and AdamW optimizers demonstrate the effectiveness of our methodology. Source code is available at https://github.com/THUDM/MoELoRA_Riemannian.

cs.LG↗

arXiv2025

Stable Soliton Microcomb Generation in X-cut Lithium Tantalate via Thermal-Assisted Photorefractive Suppression

Chip-based soliton frequency microcombs combine compact size, broad bandwidth, and high coherence, presenting a promising solution for integrated optical telecommunications, precision sensing, and spectroscopy. Recent progress in ferroelectric thin films, particularly thin-film Lithium niobate (LN) and thin-film Lithium tantalate (LT), has significantly advanced electro-optic (EO) modulation and soliton microcombs generation, leveraging their strong third-order nonlinearity and high Pockels coefficients. However, achieving soliton frequency combs in X-cut ferroelectric materials remains challenging due to the competing effects of thermo-optic and photorefractive phenomena. These issues hinder the simultaneous realization of soliton generation and high-speed EO modulation. Here, following the thermal-regulated carrier behaviour and auxiliary-laser-assisted approach, we propose a convenient mechanism to suppress both photorefractive and thermal dragging effect at once, and implement a facile method for soliton formation and its long-term stabilization in integrated X-cut LT microresonators for the first time. The resulting mode-locked states exhibit robust stability against perturbations, enabling new pathways for fully integrated photonic circuits that combine Kerr nonlinearity with high-speed EO functionality.

physics.optics↗

arXiv2025

Spatial-offset pump-probe imaging of nonradiative dynamics at optical resolution

Nonradiative photothermal (PT) and photoacoustic (PA) processes have found widespread applications in imaging, stimulation, and therapy. Mapping the generation and propagation of PA and PT waves with resolution is important to elucidate how these fields interact with biological systems. To this end, we introduce spatial offset pump-probe imaging (SOPPI). By spatially offsetting the pump beam and the probe beam, SOPPI can image simultaneously PA and PT wave propagation with nanosecond temporal resolution, micrometer spatial resolution, 65 MHz detection bandwidth, and a sensitivity of 9.9 Pa noise equivalent pressure. We first map the PA and PT evolution from a fiber emitter, and how the wave interacting with a mouse skull and brain slices. SOPPI imaging of PA waves from a tapered fiber with water as an absorber shows a wavelength-dependent generation, evanescent wave generated PA, and back-propagated acoustic Mach Cone. At last, a SOPPI-PACT is developed to reconstruct the pigment distribution inside a zebrafish larva with high precision and signal-to-noise ratio.

physics.optics↗

arXiv2025

PM-MOE: Mixture of Experts on Private Model Parameters for Personalized Federated Learning

Federated learning (FL) has gained widespread attention for its privacy-preserving and collaborative learning capabilities. Due to significant statistical heterogeneity, traditional FL struggles to generalize a shared model across diverse data domains. Personalized federated learning addresses this issue by dividing the model into a globally shared part and a locally private part, with the local model correcting representation biases introduced by the global model. Nevertheless, locally converged parameters more accurately capture domain-specific knowledge, and current methods overlook the potential benefits of these parameters. To address these limitations, we propose PM-MoE architecture. This architecture integrates a mixture of personalized modules and an energy-based personalized modules denoising, enabling each client to select beneficial personalized parameters from other clients. We applied the PM-MoE architecture to nine recent model-split-based personalized federated learning algorithms, achieving performance improvements with minimal additional training. Extensive experiments on six widely adopted datasets and two heterogeneity settings validate the effectiveness of our approach. The source code is available at \url{https://github.com/dannis97500/PM-MOE}.

cs.LG↗

arXiv2024

PowerMLP: An Efficient Version of KAN

The Kolmogorov-Arnold Network (KAN) is a new network architecture known for its high accuracy in several tasks such as function fitting and PDE solving. The superior expressive capability of KAN arises from the Kolmogorov-Arnold representation theorem and learnable spline functions. However, the computation of spline functions involves multiple iterations, which renders KAN significantly slower than MLP, thereby increasing the cost associated with model training and deployment. The authors of KAN have also noted that ``the biggest bottleneck of KANs lies in its slow training. KANs are usually 10x slower than MLPs, given the same number of parameters.'' To address this issue, we propose a novel MLP-type neural network PowerMLP that employs simpler non-iterative spline function representation, offering approximately the same training time as MLP while theoretically demonstrating stronger expressive power than KAN. Furthermore, we compare the FLOPs of KAN and PowerMLP, quantifying the faster computation speed of PowerMLP. Our comprehensive experiments demonstrate that PowerMLP generally achieves higher accuracy and a training speed about 40 times faster than KAN in various tasks.

cs.LG↗

arXiv2024

HC-SpMM: Accelerating Sparse Matrix-Matrix Multiplication for Graphs with Hybrid GPU Cores

Sparse Matrix-Matrix Multiplication (SpMM) is a fundamental operation in graph computing and analytics. However, the irregularity of real-world graphs poses significant challenges to achieving efficient SpMM operation for graph data on GPUs. Recently, significant advancements in GPU computing power and the introduction of new efficient computing cores within GPUs offer new opportunities for acceleration. In this paper, we present HC-SpMM, a pioneering algorithm that leverages hybrid GPU cores (Tensor cores and CUDA cores) to accelerate SpMM for graphs. To adapt to the computing characteristics of different GPU cores, we investigate the impact of sparse graph features on the performance of different cores, develop a data partitioning technique for the graph adjacency matrix, and devise a novel strategy for intelligently selecting the most efficient cores for processing each submatrix. Additionally, we optimize it by considering memory access and thread utilization, to utilize the computational resources to their fullest potential. To support complex graph computing workloads, we integrate HC-SpMM into the GNN training pipeline. Furthermore, we propose a kernel fusion strategy to enhance data reuse, as well as a cost-effective graph layout reorganization method to mitigate the irregular and sparse issues of real-world graphs, better fitting the computational models of hybrid GPU cores. Extensive experiments on 14 real-world graph datasets demonstrate that HC-SpMM achieves an average speedup of 1.33x and 1.23x over state-of-the-art SpMM kernels and GNN frameworks.

cs.DC↗

arXiv2024

Speech Is Not Enough: Interpreting Nonverbal Indicators of Common Knowledge and Engagement

Our goal is to develop an AI Partner that can provide support for group problem solving and social dynamics. In multi-party working group environments, multimodal analytics is crucial for identifying non-verbal interactions of group members. In conjunction with their verbal participation, this creates an holistic understanding of collaboration and engagement that provides necessary context for the AI Partner. In this demo, we illustrate our present capabilities at detecting and tracking nonverbal behavior in student task-oriented interactions in the classroom, and the implications for tracking common ground and engagement.

cs.CL↗

arXiv2024

A hybrid single quantum dot coupled cavity on a CMOS-compatible SiC photonic chip for Purcell-enhanced deterministic single-photon emission

The ability to control nonclassical light emission from a single quantum emitter by an integrated cavity may unleash new perspectives for integrated photonic quantum applications. However, coupling a single quantum emitter to cavity within photonic circuitry towards creation of the Purcell-enhanced single-photon emission is elusive due to the complexity of integrating active devices in low-loss photonic circuits. Here we demonstrate a hybrid micro-ring resonator (HMRR) coupled with self-assembled quantum dots (QDs) for cavity-enhanced deterministic single-photon emission. The HMRR cavity supports whispering-gallery modes with quality factors up to 7800. By further introducing a micro-heater, we show that the photon emission of QDs can be locally and dynamically tuned over one free spectral ranges of the HMRR (~4 nm). This allows precise tuning of individual QDs in resonance with the cavity modes, thereby enhancing single-photon emission with a Purcell factor of about 4.9. Our results on the hybrid integrated cavities coupled with two-level quantum emitters emerge as promising devices for chip-based scalable photonic quantum applications.

quant-ph↗

arXiv2024

Text2NKG: Fine-Grained N-ary Relation Extraction for N-ary relational Knowledge Graph Construction

Beyond traditional binary relational facts, n-ary relational knowledge graphs (NKGs) are comprised of n-ary relational facts containing more than two entities, which are closer to real-world facts with broader applications. However, the construction of NKGs remains at a coarse-grained level, which is always in a single schema, ignoring the order and variable arity of entities. To address these restrictions, we propose Text2NKG, a novel fine-grained n-ary relation extraction framework for n-ary relational knowledge graph construction. We introduce a span-tuple classification approach with hetero-ordered merging and output merging to accomplish fine-grained n-ary relation extraction in different arity. Furthermore, Text2NKG supports four typical NKG schemas: hyper-relational schema, event-based schema, role-based schema, and hypergraph-based schema, with high flexibility and practicality. The experimental results demonstrate that Text2NKG achieves state-of-the-art performance in F1 scores on the fine-grained n-ary relation extraction benchmark. Our code and datasets are publicly available.

cs.AI↗

arXiv2024

ChatKBQA: A Generate-then-Retrieve Framework for Knowledge Base Question Answering with Fine-tuned Large Language Models

Knowledge Base Question Answering (KBQA) aims to answer natural language questions over large-scale knowledge bases (KBs), which can be summarized into two crucial steps: knowledge retrieval and semantic parsing. However, three core challenges remain: inefficient knowledge retrieval, mistakes of retrieval adversely impacting semantic parsing, and the complexity of previous KBQA methods. To tackle these challenges, we introduce ChatKBQA, a novel and simple generate-then-retrieve KBQA framework, which proposes first generating the logical form with fine-tuned LLMs, then retrieving and replacing entities and relations with an unsupervised retrieval method, to improve both generation and retrieval more directly. Experimental results show that ChatKBQA achieves new state-of-the-art performance on standard KBQA datasets, WebQSP, and CWQ. This work can also be regarded as a new paradigm for combining LLMs with knowledge graphs (KGs) for interpretable and knowledge-required question answering. Our code is publicly available.

cs.CL↗

arXiv2024

Design of a LYSO Crystal Electromagnetic Calorimeter for DarkSHINE Experiment

This paper presents the design and optimization of a LYSO crystal electromagnetic calorimeter (ECAL) for the DarkSHINE experiment, which aims to search for dark photons as potential mediators of dark forces. The ECAL design was evaluated through comprehensive simulations, focusing on optimizing dimensions, material selection, energy distribution, and energy resolution. The ECAL configuration consists of 21$\times$21$\times$11 LYSO crystals, each measuring 2.5$\times$2.5$\times$4 cm$^3$, arranged in a staggered layout to improve signal detection efficiency. A 4 GeV energy dynamic range was established to ensure accurate energy measurements without saturation, which is essential for background rejection and signal identification. A detailed digitization model was developed to simulate the scintillation, SiPM, and ADC behaviors, providing a more realistic representation of detector performance. Additionally, the study assessed radiation damage in the ECAL region, highlighting the necessity of radiation-resistant scintillators and silicon sensors.

physics.ins-det↗

arXiv2024

Bridging Large Language Models and Graph Structure Learning Models for Robust Representation Learning

Graph representation learning, involving both node features and graph structures, is crucial for real-world applications but often encounters pervasive noise. State-of-the-art methods typically address noise by focusing separately on node features with large language models (LLMs) and on graph structures with graph structure learning models (GSLMs). In this paper, we introduce LangGSL, a robust framework that integrates the complementary strengths of pre-trained language models and GSLMs to jointly enhance both node feature and graph structure learning. In LangGSL, we first leverage LLMs to filter noise in the raw data and extract valuable cleaned information as features, enhancing the synergy of downstream models. During the mutual learning phase in LangGSL, the core idea is to leverage the relatively small language model (LM) to process local attributes and generate reliable pseudo-labels and informative node embeddings, which are then integrated into the GSLM's prediction phase. This approach enriches the global context and enhances overall performance. Meanwhile, GSLM refines the evolving graph structure constructed from the LM's output, offering updated labels back to the LM as additional guidance, thus facilitating a more effective mutual learning process. The LM and GSLM work synergistically, complementing each other's strengths and offsetting weaknesses within a variational information-maximizing framework, resulting in enhanced node features and a more robust graph structure. Extensive experiments on diverse graph datasets of varying scales and across different task scenarios demonstrate the scalability and effectiveness of the proposed approach.

cs.LG↗

arXiv2024

DIMS: Distributed Index for Similarity Search in Metric Spaces

Similarity search finds objects that are similar to a given query object based on a similarity metric. As the amount and variety of data continue to grow, similarity search in metric spaces has gained significant attention. Metric spaces can accommodate any type of data and support flexible distance metrics, making similarity search in metric spaces beneficial for many real-world applications, such as multimedia retrieval, personalized recommendation, trajectory analytics, data mining, decision planning, and distributed servers. However, existing studies mostly focus on indexing metric spaces on a single machine, which faces efficiency and scalability limitations with increasing data volume and query amount. Recent advancements in similarity search turn towards distributed methods, while they face challenges including inefficient local data management, unbalanced workload, and low concurrent search efficiency. To this end, we propose DIMS, an efficient Distributed Index for similarity search in Metric Spaces. First, we design a novel three-stage heterogeneous partition to achieve workload balance. Then, we present an effective three-stage indexing structure to efficiently manage objects. We also develop concurrent search methods with filtering and validation techniques that support efficient distributed similarity search. Additionally, we devise a cost-based optimization model to balance communication and computation cost. Extensive experiments demonstrate that DIMS significantly outperforms existing distributed similarity search approaches.

cs.DB↗

arXiv2024

Tactile Probabilistic Contact Dynamics Estimation of Unknown Objects

We study the problem of rapidly identifying contact dynamics of unknown objects in partially known environments. The key innovation of our method is a novel formulation of the contact dynamics estimation problem as the joint estimation of contact geometries and physical parameters. We leverage DeepSDF, a compact and expressive neural-network-based geometry representation over a distribution of geometries, and adopt a particle filter to estimate both the geometries in contact and the physical parameters. In addition, we couple the estimator with an active exploration strategy that plans information-gathering moves to further expedite online estimation. Through simulation and physical experiments, we show that our method estimates accurate contact dynamics with fewer than 30 exploration moves for unknown objects touching partially known environments.

cs.RO↗

arXiv2024

T2VSafetyBench: Evaluating the Safety of Text-to-Video Generative Models

The recent development of Sora leads to a new era in text-to-video (T2V) generation. Along with this comes the rising concern about its security risks. The generated videos may contain illegal or unethical content, and there is a lack of comprehensive quantitative understanding of their safety, posing a challenge to their reliability and practical deployment. Previous evaluations primarily focus on the quality of video generation. While some evaluations of text-to-image models have considered safety, they cover fewer aspects and do not address the unique temporal risk inherent in video generation. To bridge this research gap, we introduce T2VSafetyBench, a new benchmark designed for conducting safety-critical assessments of text-to-video models. We define 12 critical aspects of video generation safety and construct a malicious prompt dataset including real-world prompts, LLM-generated prompts and jailbreak attack-based prompts. Based on our evaluation results, we draw several important findings, including: 1) no single model excels in all aspects, with different models showing various strengths; 2) the correlation between GPT-4 assessments and manual reviews is generally high; 3) there is a trade-off between the usability and safety of text-to-video generative models. This indicates that as the field of video generation rapidly advances, safety risks are set to surge, highlighting the urgency of prioritizing video safety. We hope that T2VSafetyBench can provide insights for better understanding the safety of video generation in the era of generative AI.

cs.CV↗