Searcharxiv⌕ Search

arXiv subjects

Ruifeng Zhang

Publications and source records attributed to Ruifeng Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Spectral Analysis for Sparse Matrix Computation: Insights and Potential

Sparse computations are fundamental to scientific computing, graph analytics, and machine learning, yet their performance is highly sensitive to the diverse sparsity and patterns. This is because cache reuse, memory coalescing, and load balancing depend critically on the sparsity patterns. This work gives the first known exploration of the connections between sparse matrix computation and spectral analysis by treating sparse matrices as two-dimensional signals and analyzing their frequency-domain representations through Fast Fourier Transform. We show that spectral signatures uncover global structural characteristics that are not sufficiently captured by conventional spatial statistics and provide complementary information for understanding sparse computation performance. Experiments on incorporating spectral features into machine-learning-based SpMV format selection demonstrate the usefulness of such spectral analysis over a state-of-the-art spatial-only model. By uncovering the principled connections between spectral characteristics and sparse matrix computations, this work introduces a novel analytical perspective into sparse computation, and provides a new approach to enhancing the current sparse structure characterization and optimization. On pruned LLM decoding, adding spectral features improves kernel selection and yields 1.035--1.245$\times$ kernel speedups.

cs.PF↗
arXiv2026

Minimizers and Weak Solutions for Singular Born--Infeld Type Functionals

We investigate the relation between minimizers and weak solutions for a class of singular functionals arising from Born--Infeld type theories $\mathcal{L}(s)$. In the setting of an electrostatic field $s=\frac{1}{2}|\nabla\phi|^2$, $\mathcal{L}(s)$ satisfies $\lim_{s\to(1/2)^-}\mathcal{L}(s)=+\infty$, which naturally enforces the finite gradient bound $|\nabla\phi|\le 1$, also called the truncation threshold. For a prescribed extended charge density $\rho$, we consider the relation between the weak solution of the system \begin{equation} \begin{cases} -{\rm div}\left(b\left(\frac12|\nabla\phi|^2\right)\nabla\phi\right)=\rho,& \text{in }\mathbb{R}^N,\\ b(s)=\mathcal{L}'(s),\quad\lim_{s\to\frac12^-} b(s)=+\infty,\\ \lim_{|x|\to\infty}\phi(x)=0 \end{cases} \end{equation} and the minimizer $\phi_0$ of the singular functional. We propose a monotonic approximation method to handle the intrinsic singularities of $\mathcal{L}(s)$. We prove that the gradient of the minimizer never touches the singular boundary $|\nabla\phi|^2=1$; this structural result yields a key integrability property, the existence and uniqueness of the minimizer, and the corresponding variational inequality. Under the additional assumption that $\rho$ is radially distributed, we show that the minimizer is the unique weak solution. Furthermore, we establish the $C^1$ and $C^2$ regularity of the minimizer under suitable integrability conditions on $\rho$, and provide a uniform estimate for the strict spacelikeness condition $|\nabla\phi_0|\le 1-\epsilon$, where the parameter $\epsilon>0$ is explicitly characterized in terms of the spatial dimension, the spatial region, and $\rho$. Our results extend the classical Born--Infeld theory to a general class of singular Born--Infeld type theories, thereby providing a unified framework for the variational analysis and regularity of such singular functionals and systems.

math.AP↗
arXiv2026

Metric mean dimension of amenable group actions: localization and non-uniformity

In this paper, we extend Tsukamoto's recent localization formula for metric mean dimension to actions of countable discrete amenable groups, which previously applied only to $\mathbb{R}^k$- and $\mathbb{Z}^k$-actions -- by proving that the global metric mean dimension is characterized by the asymptotic entropy of pointwise $\varepsilon$-stable sets (Theorem 2.3). To achieve this generalization, we introduce equivalent definitions of the invariant using topological entropy, packing topological entropy, and Bowen's dimensional entropy, respectively. A key technical contribution is our replacement of tiling arguments with Lindenstrauss's combinatorial covering lemma, which enables us to handle the general structure of amenable groups. Furthermore, we resolve all three questions regarding uniformity raised in Section 6 of a recent paper by Yang, Chen, and Zhou by constructing counterexamples (Theorem 2.6 and Proposition 5.8), which demonstrates that the supremum and limit superior in the localization formula cannot generally be interchanged, thereby highlighting the heterogeneous nature of the convergence. These results clarify the uniformity issue and offer insights into the link between local dynamics and global invariants, while our equivalent definitions provide flexible tools for computing metric mean dimension in concrete settings.

math.DS↗
arXiv2026

Zenith: Scaling up Ranking Models for Billion-scale Livestreaming Recommendation

Accurately capturing feature interactions is essential in recommender systems, and recent trends show that scaling up model capacity could be a key driver for next-level predictive performance. While prior work has explored various model architectures to capture multi-granularity feature interactions, relatively little attention has been paid to efficient feature handling and scaling model capacity without incurring excessive inference latency. In this paper, we address this by presenting Zenith, a scalable and efficient ranking architecture that learns complex feature interactions with minimal runtime overhead. Zenith is designed to handle a few high-dimensional Prime Tokens with Token Fusion and Token Boost modules, which exhibits superior scaling laws compared to other state-of-the-art ranking methods, thanks to its improved token heterogeneity. Its real-world effectiveness is demonstrated by deploying the architecture to TikTok Live, a leading online livestreaming platform that attracts billions of users globally. Our A/B test shows that Zenith achieves +1.05%/-1.10% in online CTR AUC and Logloss, and realizes +9.93% gains in Quality Watch Session / User and +8.11% in Quality Watch Duration / User.

cs.LG↗
arXiv2025

Three-Dimensional Continuous Multi-Walled Carbon Nanotubes Network-Toughened Diamond Composite

Enhancing the fracture toughness of diamond while preserving its hardness is a significant challenge. Traditional toughening strategies have primarily focused on modulating the internal microstructural units of diamonds, including adjustments to stacking sequences, faults, nanotwinning, and the incorporation of amorphous phases, collectively referred to as intrinsic toughening. Here, we introduce an extrinsic toughening strategy to develop an unparalleled tough diamond composite with complex and abundant sp2-sp3 bonding interfaces, by incorporating highly dispersed multi-walled carbon nanotubes (MWCNTs) into the gaps of diamond grains to create a three-dimensional (3D) continuous MWCTNs network-toughen heterogeneous structure. The resultant composite exhibits a hardness of approximately 91.6 GPa and a fracture toughness of roughly 36.4 MPa.m1/2, which is six times higher than that of synthetic diamond and even surpasses that of tungsten alloys, surpassing the benefits achievable through intrinsic toughening alone. The remarkable toughening behavior can be attributed to the formation of numerous mixed sp2-sp3 bonding interactions at the 3D continuous network MWCNTs/diamond interfaces, which facilitate efficient energy dissipation. Our 3D continuous network heterogeneous structure design provides an effective approach for enhancing the fracture toughness of superhard materials, offering a new paradigm for the advanced composite ceramics.

cond-mat.other↗
arXiv2022

An Incremental Phase Mapping Approach for X-ray Diffraction Patterns using Binary Peak Representations

Despite the huge advancement in knowledge discovery and data mining techniques, the X-ray diffraction (XRD) analysis process has mostly remained untouched and still involves manual investigation, comparison, and verification. Due to the large volume of XRD samples from high-throughput XRD experiments, it has become impossible for domain scientists to process them manually. Recently, they have started leveraging standard clustering techniques, to reduce the XRD pattern representations requiring manual efforts for labeling and verification. Nevertheless, these standard clustering techniques do not handle problem-specific aspects such as peak shifting, adjacent peaks, background noise, and mixed phases; hence, resulting in incorrect composition-phase diagrams that complicate further steps. Here, we leverage data mining techniques along with domain expertise to handle these issues. In this paper, we introduce an incremental phase mapping approach based on binary peak representations using a new threshold based fuzzy dissimilarity measure. The proposed approach first applies an incremental phase computation algorithm on discrete binary peak representation of XRD samples, followed by hierarchical clustering or manual merging of similar pure phases to obtain the final composition-phase diagram. We evaluate our method on the composition space of two ternary alloy systems- Co-Ni-Ta and Co-Ti-Ta. Our results are verified by domain scientists and closely resembles the manually computed ground-truth composition-phase diagrams. The proposed approach takes us closer towards achieving the goal of complete end-to-end automated XRD analysis.

cs.LG↗
arXiv2022

Deep Learning-based Inertial Odometry for Pedestrian Tracking using Attention Mechanism and Res2Net Module

Pedestrian dead reckoning is a challenging task due to the low-cost inertial sensor error accumulation. Recent research has shown that deep learning methods can achieve impressive performance in handling this issue. In this letter, we propose inertial odometry using a deep learning-based velocity estimation method. The deep neural network based on Res2Net modules and two convolutional block attention modules is leveraged to restore the potential connection between the horizontal velocity vector and raw inertial data from a smartphone. Our network is trained using only fifty percent of the public inertial odometry dataset (RoNIN) data. Then, it is validated on the RoNIN testing dataset and another public inertial odometry dataset (OXIOD). Compared with the traditional step-length and heading system-based algorithm, our approach decreases the absolute translation error (ATE) by 76%-86%. In addition, compared with the state-of-the-art deep learning method (RoNIN), our method improves its ATE by 6%-31.4%.

cs.RO↗
arXiv2020

Time-restricted sensitivity and entropy

In this paper, we consider measure-theoretical restricted sensitivity and topological restricted sensitivities by restricting the first sensitive time. For a given topological dynamical system, we define measure-theoretical restricted asymptotic rate with respect to sensitivity, and obtain that it equal to the reciprocal of the Brin-Katok local entropy for almost every point. For topological version we have similar definitions and conclusions.

math.DS↗
arXiv2019

Anisotropic Bi2O2Se(Te) Monolayer: Realizing Ultra-High Carrier Mobility and Giant Electric Polarization in Two-Dimension

Here, we have identified the monolayer phase of Bi2O2Se as a promising two-dimensional semiconductor with ultra-high carrier mobility and giant electric polarization. Due to the strong reconstruction originated from the interlayer electrostatic force, we have applied structure prediction algorithms to explore the crystalline geometry of Bi2O2Se monolayer with the lowest total energy. Considering Se and Te belong to the same group, Bi2O2Te monolayer is also investigated based on a similar scheme. Further calculations suggest that the high carrier mobility is maintained in the monolayer phase and the moderate band gap will lead to the strong optical absorption in the visible light region. In particular, the electron mobility in Bi2O2Te can reach as high as 3610 cm2V-1s-1 at room temperature, which is almost ten times of conventional transition metal dichalcogenides (TMD) family. Because of the strong structural anisotropy, a remarkable spontaneous in-plane and out-of-plane electric polarization is additionally revealed along with significant piezoelectric properties, endowing them as promising candidates in the area of photovoltaic solar cells, optoelectronic materials and field effect transistors.

cond-mat.mtrl-sci↗
arXiv2019

Modulating Super-Exchange Strength to Achieve Robust Ferromagnetic Couplings in Two-Dimensional Semiconductors

Low-dimensional semiconducting ferromagnets have attracted considerable attention due to their promising applications as nano-size spintronics. However, realizing robust ferromagnetic couplings that can survive at high temperature is restrained by two decisive factors: super-exchange couplings and anisotropy. Despite widely explored low-dimensional anisotropy, strengthening super-exchange couplings has rarely been investigated. Here, we found that ligands with lower electronegativity can strengthen ferromagnetic super-exchange couplings and further proposed the ligand modulation strategy to enhance the Curie temperature of low-dimensional ferromagnets. Based on the metallic CrX2 (X = S, Se, Te) family, substituting ligand atoms by halides can form stable semiconducting phase as CrSeCl, CrSeBr and CrTeBr. It is interesting to discover that, the nearest ferromagnetic super-exchange couplings can be strengthened when substituting ligands from S to Se and Te. Such evolution originates from the enhanced electron hopping integral and reduced energy intervals between d and p orbits. While the second nearest anti-ferromagnetic couplings are also benefitted due to delocalized p-p interactions. Finally, ligand modulation strategy is applied in other ferromagnetic monolayers, further verifying our theory and providing a fundamental understanding on controlling super-exchange couplings in low-dimension.

cond-mat.mtrl-sci↗
arXiv2019

Discrete spectrum for amenable group actions

In this paper, we study discrete spectrum of invariant measures for countable discrete amenable group actions. We show that an invariant measure has discrete spectrum if and only if it has bounded measure complexity. We also prove that, discrete spectrum can be characterized via measure-theoretic complexity using names of a partition and the Hamming distance, and it turns out to be equivalent to both mean equicontinuity and equicontinuity in the mean.

math.DS↗
arXiv2019

Topology and Topological Sequence Entropy

Let $X$ be a compact metric space and $T:X\longrightarrow X$ be continuous. Let $h^*(T)$ be the supremum of topological sequence entropies of $T$ over all subsequences of $\mathbb Z_+$ and $S(X)$ be the set of the values $h^*(T)$ for all continuous maps $T$ on $X$. It is known that $\{0\} \subseteq S(X)\subseteq \{0, \log 2, \log 3, \ldots\}\cup \{\infty\}$. Only three possibilities for $S(X)$ have been observed so far, namely $S(X)=\{0\}$, $S(X)=\{0,\log2, \infty\}$ and $S(X)=\{0, \log 2, \log 3, \ldots\}\cup \{\infty\}$. In this paper we completely solve the problem of finding all possibilities for $S(X)$ by showing that in fact for every set $\{0\} \subseteq A \subseteq \{0, \log 2, \log 3, \ldots\}\cup \{\infty\}$ there exists a one-dimensional continuum $X_A$ with $S(X_A) = A$. In the construction of $X_A$ we use Cook continua. This is apparently the first application of these very rigid continua in dynamics. We further show that the same result is true if one considers only homeomorphisms rather than con\-ti\-nuous maps. The problem for group actions is also addressed. For some class of group actions (by homeomorphisms) we provide an analogous result, but in full generality this problem remains open. The result works also for an analogous class of semigroup actions (by continuous maps).

math.DS↗
arXiv2017

A note on dimensional entropy for amenable group actions

In this short note, for countably infinite amenable group actions, we provide topological proofs for the following results: Bowen topological entropy (dimensional entropy) of the whole space equals the usual topological entropy along tempered Følner sequences; the Hausdorff dimension of an amenable subshift (for certain metric associated to some Følner sequence) equals its topological entropy. This answers questions by Zheng and Chen (Israel Journal of Mathematics 212 (2016), 895-911) and Simpson (Theory Comput. Syst. 56 (2015), 527-543).

math.DS↗
arXiv2017

When is a dynamical system mean sensitive?

This article is devoted to study which conditions imply that a topological dynamical system is mean sensitive and which do not. Among other things we show that every uniquely ergodic, mixing system with positive entropy is mean sensitive. On the other hand we provide an example of a transitive system which is cofinitely sensitive or Devaney chaotic with positive entropy but fails to be mean sensitive. As applications of our theory and examples, we negatively answer an open question regarding equicontinuity/sensitivity dichotomies raised by Tu, we introduce and present results of locally mean equicontinuous systems and we show that mean sensitivity of the induced hyperspace does not imply that of the phase space.

math.DS↗
arXiv2017

Theoretical studies on the impact of point defect on the structures of different uranium silicides

The structures, point defects and impacts of fission products for U3Si (\b{eta}-U3Si and γ-U3Si) and USi2 (α-USi2 and \b{eta}-USi2) are studied by first-principles calculations. The lattice parameters of U3Si and USi2 are calculated and the stability of different types of point defects is predicted by their formation energies. The results show that silicon vacancies are more prone to be produced than uranium vacancies in \b{eta}-USi2 matrix, while uranium vacancies are the most stable defects of other three types of crystallographic structures. The most favorable sites of fission products (strontium, barium, cerium and neodymium) are determined in this work as well. By calculating incorporation energies of fission products, we demonstrate that the uranium site is the most favored for all the fissions products. Comparing the structural changes influenced by different fission products, it is also found that the highest volume change is caused by barium interstitials. According to the current data, rare earth elements cerium and neodymium are found to be more stable than alkaline earth metals strontium and barium in a given nuclear matrix. Finally, it is also determined that in USi2 crystal lattice fission products tend to be stabilized in uranium substitution sites, while they are likely to form precipitates from the U3Si matrix. It is expected that this work may provide new insight into the mechanism for structural evolutions of silicide nuclear fuels in a reactor.

nucl-th↗
arXiv2015

When all closed subsets are recurrent?

In the paper we study relations of rigidity, equicontinuity and pointwise recurrence between a t.d.s. $(X,T)$ and the t.d.s. $(K(X),T_K)$ induced on the hyperspace $K(X)$ of all compact subsets of $X$, and provide some characterizations. Among other examples, we construct a minimal, non-equicontinuous, distal and uniformly rigid t.d.s. and a t.d.s. which has dense small periodic sets but does not have dense distal points, solving that way open questions existing in the literature.

math.DS↗