Search arXiv⌕ Search

arXiv subjects

Zhen Zhang

Publications and source records attributed to Zhen Zhang.

At least 343 records · Page 19Linked to original sources
arXiv2024

Bayesian model averaging for nuclear symmetry energy from effective proton-neutron chemical potential difference of neutron-rich nuclei

The data-driven Bayesian model averaging is a rigorous statistical approach to combining multiple models for a unified prediction. Compared with the individual model, it provides more reliable information, especially for problems involving apparent model dependence. In this work, within both the non-relativistic Skyrme energy density functional and the nonlinear relativistic mean field model, the effective proton-neutron chemical potential difference $Δμ^*_{\rm{pn}}$ of neutron-rich nuclei is found to be strongly sensitive to the symmetry energy $E_{\rm{sym}}(ρ)$ around $2ρ_0/3$, with $ρ_0$ being the nuclear saturation density. Given discrepancies on the $Δμ^*_{\rm{pn}}$-$E_{\rm{sym}}(2ρ_0/3)$ correlations between the two models, we carry out a Bayesian model averaging analysis based on Gaussian process emulators to extract the symmetry energy around $2ρ_0/3$ from the measured $Δμ^*_{\rm{pn}}$ of 5 doubly magic nuclei $^{48}$Ca, $^{68}$Ni, $^{88}$Sr, $^{132}$Sn and $^{208}$Pb. Specifically, the $E_{\mathrm{sym}}(2ρ_0/3)$ is inferred to be $E_{\mathrm{sym}}(2ρ_0/3) = 25.6_{-1.3}^{+1.4}\,\mathrm{MeV}$ at $1σ$ confidence level. The obtained constraints on the $E_{\mathrm{sym}}(ρ)$ around $2ρ_0/3$ agree well with microscopic predictions and results from other isovector indicators.

nucl-th↗
arXiv2024

PolyTOPS: Reconfigurable and Flexible Polyhedral Scheduler

Polyhedral techniques have been widely used for automatic code optimization in low-level compilers and higher-level processes. Loop optimization is central to this technique, and several polyhedral schedulers like Feautrier, Pluto, isl and Tensor Scheduler have been proposed, each of them targeting a different architecture, parallelism model, or application scenario. The need for scenario-specific optimization is growing due to the heterogeneity of architectures. One of the most critical cases is represented by NPUs (Neural Processing Units) used for AI, which may require loop optimization with different objectives. Another factor to be considered is the framework or compiler in which polyhedral optimization takes place. Different scenarios, depending on the target architecture, compilation environment, and application domain, may require different kinds of optimization to best exploit the architecture feature set. We introduce a new configurable polyhedral scheduler, PolyTOPS, that can be adjusted to various scenarios with straightforward, high-level configurations. This scheduler allows the creation of diverse scheduling strategies that can be both scenario-specific (like state-of-the-art schedulers) and kernel-specific, breaking the concept of a one-size-fits-all scheduler approach. PolyTOPS has been used with isl and CLooG as code generators and has been integrated in MindSpore AKG deep learning compiler. Experimental results in different scenarios show good performance: a geomean speedup of 7.66x on MindSpore (for the NPU Ascend architecture) hybrid custom operators over isl scheduling, a geomean speedup up to 1.80x on PolyBench on different multicore architectures over Pluto scheduling. Finally, some comparisons with different state-of-the-art tools are presented in the PolyMage scenario.

cs.DC↗
arXiv2024

A Video Coding Method Based on Neural Network for CLIC2024

This paper presents a video coding scheme that combines traditional optimization methods with deep learning methods based on the Enhanced Compression Model (ECM). In this paper, the traditional optimization methods adaptively adjust the quantization parameter (QP). The key frame QP offset is set according to the video content characteristics, and the coding tree unit (CTU) level QP of all frames is also adjusted according to the spatial-temporal perception information. Block importance mapping technology (BIM) is also introduced, which adjusts the QP according to the block importance. Meanwhile, the deep learning methods propose a convolutional neural network-based loop filter (CNNLF), which is turned on/off based on the rate-distortion optimization at the CTU and frame level. Besides, intra-prediction using neural networks (NN-intra) is proposed to further improve compression quality, where 8 neural networks are used for predicting blocks of different sizes. The experimental results show that compared with ECM-3.0, the proposed traditional methods and adding deep learning methods improve the PSNR by 0.54 dB and 1 dB at 0.05Mbps, respectively; 0.38 dB and 0.71dB at 0.5 Mbps, respectively, which proves the superiority of our method.

eess.IV↗
arXiv2024

GBSS:a global building semantic segmentation dataset for large-scale remote sensing building extraction

Semantic segmentation techniques for extracting building footprints from high-resolution remote sensing images have been widely used in many fields such as urban planning. However, large-scale building extraction demands higher diversity in training samples. In this paper, we construct a Global Building Semantic Segmentation (GBSS) dataset (The dataset will be released), which comprises 116.9k pairs of samples (about 742k buildings) from six continents. There are significant variations of building samples in terms of size and style, so the dataset can be a more challenging benchmark for evaluating the generalization and robustness of building semantic segmentation models. We validated through quantitative and qualitative comparisons between different datasets, and further confirmed the potential application in the field of transfer learning by conducting experiments on subsets.

cs.CV↗
arXiv2023

Slapo: A Schedule Language for Progressive Optimization of Large Deep Learning Model Training

Recent years have seen an increase in the development of large deep learning (DL) models, which makes training efficiency crucial. Common practice is struggling with the trade-off between usability and performance. On one hand, DL frameworks such as PyTorch use dynamic graphs to facilitate model developers at a price of sub-optimal model training performance. On the other hand, practitioners propose various approaches to improving the training efficiency by sacrificing some of the flexibility, ranging from making the graph static for more thorough optimization (e.g., XLA) to customizing optimization towards large-scale distributed training (e.g., DeepSpeed and Megatron-LM). In this paper, we aim to address the tension between usability and training efficiency through separation of concerns. Inspired by DL compilers that decouple the platform-specific optimizations of a tensor-level operator from its arithmetic definition, this paper proposes a schedule language, Slapo, to decouple model execution from definition. Specifically, Slapo works on a PyTorch model and uses a set of schedule primitives to convert the model for common model training optimizations such as high-performance kernels, effective 3D parallelism, and efficient activation checkpointing. Compared to existing optimization solutions, Slapo progressively optimizes the model "as-needed" through high-level primitives, and thus preserving programmability and debuggability for users to a large extent. Our evaluation results show that by scheduling the existing hand-crafted optimizations in a systematic way using Slapo, we are able to improve training throughput by up to 2.92x on a single machine with 8 NVIDIA V100 GPUs, and by up to 1.41x on multiple machines with up to 64 GPUs, when compared to the out-of-the-box performance of DeepSpeed and Megatron-LM.

cs.LG↗
arXiv2023

Coreference Graph Guidance for Mind-Map Generation

Mind-map generation aims to process a document into a hierarchical structure to show its central idea and branches. Such a manner is more conducive to understanding the logic and semantics of the document than plain text. Recently, a state-of-the-art method encodes the sentences of a document sequentially and converts them to a relation graph via sequence-to-graph. Though this method is efficient to generate mind-maps in parallel, its mechanism focuses more on sequential features while hardly capturing structural information. Moreover, it's difficult to model long-range semantic relations. In this work, we propose a coreference-guided mind-map generation network (CMGN) to incorporate external structure knowledge. Specifically, we construct a coreference graph based on the coreference semantic relationship to introduce the graph structure information. Then we employ a coreference graph encoder to mine the potential governing relations between sentences. In order to exclude noise and better utilize the information of the coreference graph, we adopt a graph enhancement module in a contrastive learning manner. Experimental results demonstrate that our model outperforms all the existing methods. The case study further proves that our model can more accurately and concisely reveal the structure and semantics of a document. Code and data are available at https://github.com/Cyno2232/CMGN.

cs.CL↗
arXiv2023

Oobleck: Resilient Distributed Training of Large Models Using Pipeline Templates

Oobleck enables resilient distributed training of large DNN models with guaranteed fault tolerance. It takes a planning-execution co-design approach, where it first generates a set of heterogeneous pipeline templates and instantiates at least $f+1$ logically equivalent pipeline replicas to tolerate any $f$ simultaneous failures. During execution, it relies on already-replicated model states across the replicas to provide fast recovery. Oobleck provably guarantees that some combination of the initially created pipeline templates can be used to cover all available resources after $f$ or fewer simultaneous failures, thereby avoiding resource idling at all times. Evaluation on large DNN models with billions of parameters shows that Oobleck provides consistently high throughput, and it outperforms state-of-the-art fault tolerance solutions like Bamboo and Varuna by up to $29.6x$.

cs.DC↗
arXiv2023

An operator-splitting optimization approach for phase-field simulation of equilibrium shapes of crystals

Computing equilibrium shapes of crystals (ESC) is a challenging problem in materials science that involves minimizing an orientation-dependent (i.e., anisotropic) surface energy functional subject to a prescribed mass constraint. The highly nonlinear and singular anisotropic terms in the problem make it very challenging from both the analytical and numerical aspects. Especially, when the strength of anisotropy is very strong (i.e., strongly anisotropic cases), the ESC will form some singular, sharp corners even if the surface energy function is smooth. Traditional numerical approaches, such as the $H^{-1}$ gradient flow, are unable to produce true sharp corners due to the necessary addition of a high-order regularization term that penalizes sharp corners and rounds them off. In this paper, we propose a new numerical method based on the Davis-Yin splitting (DYS) optimization algorithm to predict the ESC instead of using gradient flow approaches. We discretize the infinite-dimensional phase-field energy functional in the absence of regularization terms and transform it into a finite-dimensional constraint minimization problem. The resulting optimization problem is solved using the DYS method which automatically guarantees the mass-conservation and bound-preserving properties. We also prove the global convergence of the proposed algorithm. These desired properties are numerically observed. In particular, the proposed method can produce real sharp corners with satisfactory accuracy. Finally, we present numerous numerical results to demonstrate that the ESC can be well simulated under different types of anisotropic surface energies, which also confirms the effectiveness and efficiency of the proposed method.

math.NA↗
arXiv2023

BERT4ETH: A Pre-trained Transformer for Ethereum Fraud Detection

As various forms of fraud proliferate on Ethereum, it is imperative to safeguard against these malicious activities to protect susceptible users from being victimized. While current studies solely rely on graph-based fraud detection approaches, it is argued that they may not be well-suited for dealing with highly repetitive, skew-distributed and heterogeneous Ethereum transactions. To address these challenges, we propose BERT4ETH, a universal pre-trained Transformer encoder that serves as an account representation extractor for detecting various fraud behaviors on Ethereum. BERT4ETH features the superior modeling capability of Transformer to capture the dynamic sequential patterns inherent in Ethereum transactions, and addresses the challenges of pre-training a BERT model for Ethereum with three practical and effective strategies, namely repetitiveness reduction, skew alleviation and heterogeneity modeling. Our empirical evaluation demonstrates that BERT4ETH outperforms state-of-the-art methods with significant enhancements in terms of the phishing account detection and de-anonymization tasks. The code for BERT4ETH is available at: https://github.com/git-disl/BERT4ETH.

cs.CR↗
arXiv2023

Parameter-Efficient Cross-lingual Transfer of Vision and Language Models via Translation-based Alignment

Pre-trained vision and language models such as CLIP have witnessed remarkable success in connecting images and texts with a primary focus on English texts. Despite recent efforts to extend CLIP to support other languages, disparities in performance among different languages have been observed due to uneven resource availability. Additionally, current cross-lingual transfer methods of those pre-trained models would consume excessive resources for a large number of languages. Therefore, we propose a new parameter-efficient cross-lingual transfer learning framework that utilizes a translation-based alignment method to mitigate multilingual disparities and explores parameter-efficient fine-tuning methods for parameter-efficient cross-lingual transfer. Extensive experiments on XTD and Multi30K datasets, covering 11 languages under zero-shot, few-shot, and full-dataset learning scenarios, show that our framework significantly reduces the multilingual disparities among languages and improves cross-lingual transfer results, especially in low-resource scenarios, while only keeping and fine-tuning an extremely small number of parameters compared to the full model (e.g., Our framework only requires 0.16\% additional parameters of a full-model for each language in the few-shot learning scenario). The codes are available at \url{https://github.com/eric-ai-lab/PECTVLM}. The codes are available at \url{https://github.com/eric-ai-lab/PECTVLM}.

cs.CL↗
arXiv2023

Central Speed of Sound, Trace Anomaly and Observables of Neutron Stars from Perturbative Analyses of Scaled TOV Equations

The central speed of sound (SS) measures the stiffness of the Equation of State (EOS) of superdense neutron star (NS) matter. Its variations with density and radial coordinate in NSs in conventional analyses often suffer from uncertainties of the specific nuclear EOSs used. Using the central SS and NS mass/radius scaling obtained from solving perturbatively the scaled Tolman-Oppenheimer-Volkoff (TOV) equations, we study the variations of SS, trace anomaly and several closely related properties of NSs in an EOS-model independent manner. We find that the SS increases with the reduced central pressure $\widehat{P}_{\rm{c}}\equiv P_{\rm{c}}/\varepsilon_{\rm{c}}$ (scaled by the central energy density $\varepsilon_{\rm{c}}$), and the conformal bound for SS tends to break down for NSs with masses higher than about 1.9$M_{\odot}$. The ratio $P/\varepsilon$ is upper bounded as $P/\varepsilon\lesssim0.374$ around the centers of stable NSs. We demonstrate that it is an intrinsic property of strong-field gravity and is more relevant than the perturbative QCD bound on it. While a sharp phase transition at high densities characterized by a sudden vanishing of SS in cores of massive NSs are basically excluded, the probability for a continuous crossover signaled by a peaked radial profile of SS is found to be enhanced as $\widehat{P}_{\rm{c}}$ decreases, implying it likely happens near the centers of massive NSs. Moreover, a new and more stringent causality boundary as $R_{\max}/\rm{km}\gtrsim 4.73M_{\rm{NS}}^{\max}/M_{\odot}+1.14$ for NS M-R curve is found to be excellently consistent with observational data on NS masses and radii. Furthermore, new constraints on the ultimate energy density and pressure allowed in NSs before collapsing into black holes are obtained and compared with earlier predictions in the literature.

nucl-th↗
arXiv2023

Live Graph Lab: Towards Open, Dynamic and Real Transaction Graphs with NFT

Numerous studies have been conducted to investigate the properties of large-scale temporal graphs. Despite the ubiquity of these graphs in real-world scenarios, it's usually impractical for us to obtain the whole real-time graphs due to privacy concerns and technical limitations. In this paper, we introduce the concept of {\it Live Graph Lab} for temporal graphs, which enables open, dynamic and real transaction graphs from blockchains. Among them, Non-fungible tokens (NFTs) have become one of the most prominent parts of blockchain over the past several years. With more than \$40 billion market capitalization, this decentralized ecosystem produces massive, anonymous and real transaction activities, which naturally forms a complicated transaction network. However, there is limited understanding about the characteristics of this emerging NFT ecosystem from a temporal graph analysis perspective. To mitigate this gap, we instantiate a live graph with NFT transaction network and investigate its dynamics to provide new observations and insights. Specifically, through downloading and parsing the NFT transaction activities, we obtain a temporal graph with more than 4.5 million nodes and 124 million edges. Then, a series of measurements are presented to understand the properties of the NFT ecosystem. Through comparisons with social, citation, and web networks, our analyses give intriguing findings and point out potential directions for future exploration. Finally, we also study machine learning models in this live graph to enrich the current datasets and provide new opportunities for the graph community. The source codes and dataset are available at https://livegraphlab.github.io.

cs.AI↗
arXiv2023

Quantum anomalous Hall state in a fluorinated 1T-MoSe$_2$ monolayer

The quantum anomalous Hall state with a large band gap and a high Chern number is significant for practical applications in spintronics. By performing first-principles calculations, we investigate electronic properties of the fully fluorinated 1T-MoSe$_{2}$ monolayer. Without considering the spin-orbit coupling, the band structure demonstrates single-spin semi-metallic properties and the trigonal warping around $K_{\pm}$ valleys. The introduction of the spin-orbit coupling opens considerable band gaps of $117.2$ meV around the two valleys, leading to a nontrivial quantum anomalous Hall state with a Chern number of $|C|=2$, which provides two chiral dissipationless transport channels from topological edge states and associated quantized anomalous Hall conductivity. In addition, an effective model is constructed to describe the low-energy physics of the monolayer. Our findings in the MoSe$_{2}$F$_{2}$ monolayer sheds light on large-gap quantum anomalous Hall states in two-dimensional materials with the chemical functionalization, and provides opportunities in designing low-power and noise-tolerant spintronic devices.

cond-mat.mes-hall↗
arXiv2023

RobustFair: Adversarial Evaluation through Fairness Confusion Directed Gradient Search

Deep neural networks (DNNs) often face challenges due to their vulnerability to various adversarial perturbations, including false perturbations that undermine prediction accuracy and biased perturbations that cause biased predictions for similar inputs. This paper introduces a novel approach, RobustFair, to evaluate the accurate fairness of DNNs when subjected to these false or biased perturbations. RobustFair employs the notion of the fairness confusion matrix induced in accurate fairness to identify the crucial input features for perturbations. This matrix categorizes predictions as true fair, true biased, false fair, and false biased, and the perturbations guided by it can produce a dual impact on instances and their similar counterparts to either undermine prediction accuracy (robustness) or cause biased predictions (individual fairness). RobustFair then infers the ground truth of these generated adversarial instances based on their loss function values approximated by the total derivative. To leverage the generated instances for trustworthiness improvement, RobustFair further proposes a data augmentation strategy to prioritize adversarial instances resembling the original training set, for data augmentation and model retraining. Notably, RobustFair excels at detecting intertwined issues of robustness and individual fairness, which are frequently overlooked in standard robustness and individual fairness evaluations. This capability empowers RobustFair to enhance both robustness and individual fairness evaluations by concurrently identifying defects in either domain. Empirical case studies and quantile regression analyses on benchmark datasets demonstrate the effectiveness of the fairness confusion matrix guided perturbation for false or biased adversarial instance generation.

cs.LG↗
arXiv2023

RADE: Reference-Assisted Dialogue Evaluation for Open-Domain Dialogue

Evaluating open-domain dialogue systems is challenging for reasons such as the one-to-many problem, i.e., many appropriate responses other than just the golden response. As of now, automatic evaluation methods need better consistency with humans, while reliable human evaluation can be time- and cost-intensive. To this end, we propose the Reference-Assisted Dialogue Evaluation (RADE) approach under the multi-task learning framework, which leverages the pre-created utterance as reference other than the gold response to relief the one-to-many problem. Specifically, RADE explicitly compares reference and the candidate response to predict their overall scores. Moreover, an auxiliary response generation task enhances prediction via a shared encoder. To support RADE, we extend three datasets with additional rated responses other than just a golden response by human annotation. Experiments on our three datasets and two existing benchmarks demonstrate the effectiveness of our method, where Pearson, Spearman, and Kendall correlations with human evaluation outperform state-of-the-art baselines.

cs.CL↗
arXiv2023

Prediction of Van Hove singularity systems in ternary borides

A computational search for stable structures among both $α$ and $β$ phases of ternary ATB4 borides (A= Mg, Ca, Sr, Ba, Al, Ga, and Zn, T is 3d or 4d transition elements) has been performed. We found that $α$-ATB4 compounds with A=Mg, Ca, Al, and T=V, Cr, Mn, Fe, Ni, and Co form a family of structurally stable or almost stable materials. These systems are metallic in non-magnetic states and characterized by the formation of the localized molecular-like state of 3d transition metal atom dimers, which leads to the appearance of numerous Van Hove singularities (VHS) in the electronic spectrum. The closeness of these VHS to the Fermi level can be easily tuned by electron doping. For the atoms in the middle of the 3d row (Cr, Mn, and Fe), these VHS led to magnetic instabilities and new magnetic ground states with a weakly metallic or semiconducting nature. The magnetic ground states in these systems appear as an analog of the spin glass state. Experimental attempts to produce MgFeB4 and associated challenges are discussed, and promising directions for further synthetic studies are formulated.

cond-mat.mtrl-sci↗
arXiv2023

Dark Energy on Astrophysical Scales and Its Detection in the Milky Way

The origin and nature of dark energy is one of the most significant challenges in modern science. This research aims to investigate dark energy on astrophysical scales and provide a cosmology-independent method to measure its equation-of-state parameter $w$. To accomplish this, we introduce the concept of a perfect fluid in any static, curved spacetime, and express the energy-momentum tensor of the perfect fluid in a general isotropic form, namely Weinberg's isotropic form. This enables us to define an equation-of-state parameter in a physical and global manner. Within this theoretical framework, we demonstrate that the energy-momentum tensor of dark energy on different scales can take the general isotropic form. Furthermore, we explore the SdS$_{w}$ spacetime and establish its connection with dark energy in cosmology through the equation-of-state parameter $w$. In the SdS$_{w}$ spacetime, a repulsive dark force can be induced by dark energy locally. We then apply the concept of the dark force to realistic astrophysical systems using the Poisson equation. Finally, we find that an anomaly in the Milky Way rotation curve can be quantitatively interpreted by the dark force. By fitting the galactic curve, we are able to obtain the value of the equation-of-state parameter of dark energy, independently of specific dark energy models.

gr-qc↗
arXiv2023

Depth analysis of battery performance based on a data-driven approach

Capacity attenuation is one of the most intractable issues in the current of application of the cells. The disintegration mechanism is well known to be very complex across the system. It is a great challenge to fully comprehend this process and predict the process accurately. Thus, the machine learning (ML) technology is employed to predict the specific capacity change of the cell throughout the cycle and grasp this intricate procedure. Different from the previous work, according to the WOA-ELM model proposed in this work (R2 = 0.9999871), the key factors affecting the specific capacity of the battery are determined, and the defects in the machine learning black box are overcome by the interpretable model. Their connection with the structural damage of electrode materials and battery failure during battery cycling is comprehensively explained, revealing their essentiality to battery performance, which is conducive to superior research on contemporary batteries and modification.

cs.AI↗