arXiv subjects

Gang Chen

Publications and source records attributed to Gang Chen.

At least 235 records · Page 13Linked to original sources

arXiv2024

Higher-order Klein bottle topological insulator in three-dimensional acoustic crystals

Topological phases of matter are classified based on symmetries, with nonsymmorphic symmetries like glide reflections and screw rotations being of particular importance in the classification. In contrast to extensively studied glide reflections in real space, introducing space-dependent gauge transformations can lead to momentum-space glide reflection symmetries, which may even change the fundamental domain for topological classifications, e.g., from a torus to a Klein bottle. Here we discover a new class of three-dimensional (3D) higher-order topological insulators, protected by a pair of momentum-space glide reflections. It supports gapless hinge modes, as dictated by the quadrupole moment and Wannier Hamiltonians defined on a Klein bottle manifold, and we introduce two topological invariants to characterize this phase. Our predicted topological hinge modes are experimentally verified in a 3D-printed acoustic crystal, providing direct evidence for 3D higher-order Klein bottle topological insulators. Our results not only showcase the remarkable role of momentum-space glide reflections in topological classifications, but also pave the way for experimentally exploring physical effects arising from momentum-space nonsymmorphic symmetries.

cond-mat.mes-hall↗

arXiv2024

Pre-Trained Model Recommendation for Downstream Fine-tuning

As a fundamental problem in transfer learning, model selection aims to rank off-the-shelf pre-trained models and select the most suitable one for the new target task. Existing model selection techniques are often constrained in their scope and tend to overlook the nuanced relationships between models and tasks. In this paper, we present a pragmatic framework \textbf{Fennec}, delving into a diverse, large-scale model repository while meticulously considering the intricate connections between tasks and models. The key insight is to map all models and historical tasks into a transfer-related subspace, where the distance between model vectors and task vectors represents the magnitude of transferability. A large vision model, as a proxy, infers a new task's representation in the transfer space, thereby circumventing the computational burden of extensive forward passes. We also investigate the impact of the inherent inductive bias of models on transfer results and propose a novel method called \textbf{archi2vec} to encode the intricate structures of models. The transfer score is computed through straightforward vector arithmetic with a time complexity of $\mathcal{O}(1)$. Finally, we make a substantial contribution to the field by releasing a comprehensive benchmark. We validate the effectiveness of our framework through rigorous testing on two benchmarks. The benchmark and the code will be publicly available in the near future.

cs.CV↗

arXiv2024

Kinematic Hopf algebra and BCJ numerators at finite $α'$

In this letter, starting from a kinematic Hopf algebra, we first construct a closed-form formula for all Bern-Carrasco-Johansson (BCJ) numerators in Yang-Mills (YM) theory with infinite orders of $α'$ corrections, known as $\rm DF^2+YM$ theory, when coupled to two heavy particles which can be removed through a simple factorization limit. The full $α'$ dependence appears simply in massive physical propagator factors, with factorization strongly constraining the construction. The intricate structure induced by the massive poles also naturally leads us to find a novel closed-form and local expression for BCJ numerators in usual pure YM theory, based directly on the kinematic Hopf algebra.

hep-th↗

arXiv2024

FL-GUARD: A Holistic Framework for Run-Time Detection and Recovery of Negative Federated Learning

Federated learning (FL) is a promising approach for learning a model from data distributed on massive clients without exposing data privacy. It works effectively in the ideal federation where clients share homogeneous data distribution and learning behavior. However, FL may fail to function appropriately when the federation is not ideal, amid an unhealthy state called Negative Federated Learning (NFL), in which most clients gain no benefit from participating in FL. Many studies have tried to address NFL. However, their solutions either (1) predetermine to prevent NFL in the entire learning life-cycle or (2) tackle NFL in the aftermath of numerous learning rounds. Thus, they either (1) indiscriminately incur extra costs even if FL can perform well without such costs or (2) waste numerous learning rounds. Additionally, none of the previous work takes into account the clients who may be unwilling/unable to follow the proposed NFL solutions when using those solutions to upgrade an FL system in use. This paper introduces FL-GUARD, a holistic framework that can be employed on any FL system for tackling NFL in a run-time paradigm. That is, to dynamically detect NFL at the early stage (tens of rounds) of learning and then to activate recovery measures when necessary. Specifically, we devise a cost-effective NFL detection mechanism, which relies on an estimation of performance gain on clients. Only when NFL is detected, we activate the NFL recovery process, in which each client learns in parallel an adapted model when training the global model. Extensive experiment results confirm the effectiveness of FL-GUARD in detecting NFL and recovering from NFL to a healthy learning state. We also show that FL-GUARD is compatible with previous NFL solutions and robust against clients unwilling/unable to take any recovery measures.

cs.LG↗

arXiv2024

Resummed spinning waveforms from five-point amplitudes

We compute the classical tree-level five-point amplitude for the two-to-two scattering of spinning celestial objects with the emission of a graviton. Using this five-point amplitude, we then turn to the computation of the leading-order time-domain gravitational waveform. The method we describe is suitable for arbitrary values of classical spin of Kerr black holes and does not require any expansion in powers of the spin. In this paper we illustrate it in the simpler case of the scattering of one Kerr and one Schwarzschild black hole. An important ingredient of our calculation is a novel form of the Compton amplitude with spinning particles including contact terms derived from matching to black-hole perturbation theory calculations. This ensures that our waveform is valid up to at least fourth order in the spin. Our method can be applied immediately to generate improved waveforms once higher-order contact terms in the Compton amplitude become available. Finally, we show the formula for the gravitational memory to all orders in the spin, which is in agreement with our results.

hep-th↗

arXiv2024

Exact Numerical Solution of Stochastic Master Equations for Conditional Spin Squeezing

Stochastic master equations are often used to describe conditional spin squeezing of atomic ensemble, but are limited so far to the systems with few atoms due to the exponentially increased Hilbert space. In this article, we present an exact numerical solution of these equations for systems with identical atoms by mapping identical density matrix elements to a single quantity characterized by collective quantum numbers, and apply it to the system with hundred atoms in a bad cavity subject to a homodyne detection. We demonstrate that the spin squeezing can be vividly illustrated by the Gaussian-like distribution of the collective density matrix elements, and we examine the influence of the probe field strength and polarization, the detection efficiency, the spontaneous emission rate and the number of atoms. Our exact approach can play an important role in gauging the approximate approaches applied for systems with more atoms, such as Gaussian-state formalism and stochastic mean-field approach, and it permits also exploration of entanglement effects beyond these approaches.

quant-ph↗

arXiv2024

Towards Cross-Table Masked Pretraining for Web Data Mining

Tabular data pervades the landscape of the World Wide Web, playing a foundational role in the digital architecture that underpins online information. Given the recent influence of large-scale pretrained models like ChatGPT and SAM across various domains, exploring the application of pretraining techniques for mining tabular data on the web has emerged as a highly promising research direction. Indeed, there have been some recent works around this topic where most (if not all) of them are limited in the scope of a fixed-schema/single table. Due to the scale of the dataset and the parameter size of the prior models, we believe that we have not reached the ''BERT moment'' for the ubiquitous tabular data. The development on this line significantly lags behind the counterpart research domains such as natural language processing. In this work, we first identify the crucial challenges behind tabular data pretraining, particularly overcoming the cross-table hurdle. As a pioneering endeavor, this work mainly (i)-contributes a high-quality real-world tabular dataset, (ii)-proposes an innovative, generic, and efficient cross-table pretraining framework, dubbed as CM2, where the core to it comprises a semantic-aware tabular neural network that uniformly encodes heterogeneous tables without much restriction and (iii)-introduces a novel pretraining objective -- prompt Masked Table Modeling (pMTM) -- inspired by NLP but intricately tailored to scalable pretraining on tables. Our extensive experiments demonstrate CM2's state-of-the-art performance and validate that cross-table pretraining can enhance various downstream tasks.

cs.LG↗

arXiv2024

Schur rings over Free Abelian Group of Rank Two

Schur rings are a type of subrings of group rings afforded by a partition of the underlined group. In this paper, Schur rings over free abelian group of rank two are classified under the assumption that one of the direct factor is a union of some basic sets. There are eight different types, and all but one type of which are traditional.

math.GR↗

arXiv2024

Evaluating and Enhancing Large Language Models Performance in Domain-specific Medicine: Osteoarthritis Management with DocOA

The efficacy of large language models (LLMs) in domain-specific medicine, particularly for managing complex diseases such as osteoarthritis (OA), remains largely unexplored. This study focused on evaluating and enhancing the clinical capabilities of LLMs in specific domains, using osteoarthritis (OA) management as a case study. A domain specific benchmark framework was developed, which evaluate LLMs across a spectrum from domain-specific knowledge to clinical applications in real-world clinical scenarios. DocOA, a specialized LLM tailored for OA management that integrates retrieval-augmented generation (RAG) and instruction prompts, was developed. The study compared the performance of GPT-3.5, GPT-4, and a specialized assistant, DocOA, using objective and human evaluations. Results showed that general LLMs like GPT-3.5 and GPT-4 were less effective in the specialized domain of OA management, particularly in providing personalized treatment recommendations. However, DocOA showed significant improvements. This study introduces a novel benchmark framework which assesses the domain-specific abilities of LLMs in multiple aspects, highlights the limitations of generalized LLMs in clinical contexts, and demonstrates the potential of tailored approaches for developing domain-specific medical LLMs.

cs.CL↗

arXiv2024

CARAT: Contrastive Feature Reconstruction and Aggregation for Multi-Modal Multi-Label Emotion Recognition

Multi-modal multi-label emotion recognition (MMER) aims to identify relevant emotions from multiple modalities. The challenge of MMER is how to effectively capture discriminative features for multiple labels from heterogeneous data. Recent studies are mainly devoted to exploring various fusion strategies to integrate multi-modal information into a unified representation for all labels. However, such a learning scheme not only overlooks the specificity of each modality but also fails to capture individual discriminative features for different labels. Moreover, dependencies of labels and modalities cannot be effectively modeled. To address these issues, this paper presents ContrAstive feature Reconstruction and AggregaTion (CARAT) for the MMER task. Specifically, we devise a reconstruction-based fusion mechanism to better model fine-grained modality-to-label dependencies by contrastively learning modal-separated and label-specific features. To further exploit the modality complementarity, we introduce a shuffle-based aggregation strategy to enrich co-occurrence collaboration among labels. Experiments on two benchmark datasets CMU-MOSEI and M3ED demonstrate the effectiveness of CARAT over state-of-the-art methods. Code is available at https://github.com/chengzju/CARAT.

cs.MM↗

arXiv2024

Theoretical Study on Superradiant Raman Scattering with Rubidium Atoms in An Optical Cavity

Superradiant Raman scattering of Rubidium atoms has been explored in the experiment [Nature 484, 78 (2012)] to prove the concept of the superradiant laser, which attracts significant attentions in quantum metrology due to the expected ultra-narrow linewidth down to millihertz. To better understand the physics involved in this experiment, we have developed a quantum master equation theory by treating the Rubidium atoms as three-level systems, and coupling them with a dressed laser and an optical cavity. Our simulations show different superradiant Raman scattering pulses for the systems within the crossover and strong coupling regime, and the shifted and broader spectrum of the steady-state Raman scattering. Thus, our studies provide a unified view on the superradiant Raman scattering pulses, and an alternative explanation to the broad spectrum of the steady-state Raman scattering, as observed in the experiment. In future, our theory can be readily applied to study other interesting phenomena relying on the superradiant Raman scattering, such as magnetic field sensing, real-time tracking of quantum phase, Dicke phase transition of non-equilibrium dynamics and so on.

quant-ph↗

arXiv2023

Deep Radon Prior: A Fully Unsupervised Framework for Sparse-View CT Reconstruction

Although sparse-view computed tomography (CT) has significantly reduced radiation dose, it also introduces severe artifacts which degrade the image quality. In recent years, deep learning-based methods for inverse problems have made remarkable progress and have become increasingly popular in CT reconstruction. However, most of these methods suffer several limitations: dependence on high-quality training data, weak interpretability, etc. In this study, we propose a fully unsupervised framework called Deep Radon Prior (DRP), inspired by Deep Image Prior (DIP), to address the aforementioned limitations. DRP introduces a neural network as an implicit prior into the iterative method, thereby realizing cross-domain gradient feedback. During the reconstruction process, the neural network is progressively optimized in multiple stages to narrow the solution space in radon domain for the under-constrained imaging protocol, and the convergence of the proposed method has been discussed in this work. Compared with the popular pre-trained method, the proposed framework requires no dataset and exhibits superior interpretability and generalization ability. The experimental results demonstrate that the proposed method can generate detailed images while effectively suppressing image artifacts.Meanwhile, DRP achieves comparable or better performance than the supervised methods.

eess.IV↗

arXiv2023

FreeAL: Towards Human-Free Active Learning in the Era of Large Language Models

Collecting high-quality labeled data for model training is notoriously time-consuming and labor-intensive for various NLP tasks. While copious solutions, such as active learning for small language models (SLMs) and prevalent in-context learning in the era of large language models (LLMs), have been proposed and alleviate the labeling burden to some extent, their performances are still subject to human intervention. It is still underexplored how to reduce the annotation cost in the LLMs era. To bridge this, we revolutionize traditional active learning and propose an innovative collaborative learning framework FreeAL to interactively distill and filter the task-specific knowledge from LLMs. During collaborative training, an LLM serves as an active annotator inculcating its coarse-grained knowledge, while a downstream SLM is incurred as a student to filter out high-quality in-context samples to feedback LLM for the subsequent label refinery. Extensive experiments on eight benchmark datasets demonstrate that FreeAL largely enhances the zero-shot performances for both SLM and LLM without any human supervision. The code is available at https://github.com/Justherozen/FreeAL .

cs.CL↗

arXiv2023

Optimal and Heuristic Min-Reg Scheduling Algorithms for GPU Programs

Given a basic block of instructions, finding a schedule that requires the minimum number of registers for evaluation is a well-known problem. The problem is NP-complete when the dependences among instructions form a directed-acyclic graph instead of a tree. We are striving to find efficient approximation algorithms for this problem not simply because it is an interesting graph optimization problem in theory. A good solution to this problem is also an essential component in solving the more complex instruction scheduling problem on GPU. In this paper, we start with explanations on why this problem is important in GPU instruction scheduling. We then explore two different approaches to tackling this problem. First we model this problem as a constraint-programming problem. Using a state-of-the-art CP-SAT solver, we can find optimal answers for much larger cases than previous works on a modest desktop PC. Second, guided by the optimal answers, we design and evaluate heuristics that can be applied to the polynomial-time list scheduling algorithms. A combination of those heuristics can achieve the register-pressure results that are about 17\% higher than the optimal minimum on average. However, there are still near 6\% cases in which the register pressure by the heuristic approach is 50\% higher than the optimal minimum.

cs.PL↗

arXiv2023

Infinite critical boson non-Fermi liquid on heterostructure interfaces

We study the emergence of non-Fermi liquid on heterostructure interfaces where there exists an infinite number of critical boson modes accounting for the magnetic fluctuations in two spatial dimensions. The interfacial Dzyaloshinskii-Moriya interaction naturally arises in magnetic interactions due to the absence of inversion symmetry, resulting in a degenerate contour for the low-energy bosonic modes in the momentum space which simultaneously becomes critical near the magnetic phase transition. The itinerant electrons are scattered by the critical boson contour via the Yukawa coupling. When the boson contour is much smaller than the Fermi surface, it is shown that, there exists a regime with a dynamic critical exponent z=3 while the boson contour still controls the low-energy magnetic fluctuations. Using a self-consistent renormalization calculation for this regime, we uncover a prominent non-Fermi liquid behavior in the resistivity with a characteristic temperature scaling power. These findings open up new avenues for understanding boson-fermion interactions and the novel fermionic quantum criticality.

cond-mat.str-el↗

arXiv2023

On the Weisfeiler algorithm of depth-$1$ stabilization

An origin of the multidimensional Weisfeiler-Leman algorithm goes back to a refinement procedure of deep stabilization, introduced by B. Weisfeiler in a paper included in the collective monograph ``On construction and identification of graphs"(1976). This procedure is recursive and the recursion starts from an algorithm of depth-$1$ stabilization, which has never been discussed in the literature. A goal of the present paper is to show that a simplified algorithm of the depth-$1$ stabilization has the same power as the $3$-dimensional Weisfeiler-Leman algorithm. It is proved that the class of coherent configurations obtained at the output of this simplified algorithm coincides with the class introduced earlier by the third author. As an application we also prove that if there exist at least two nonisomorphic projective planes of order $q$, then the Weisfeiler-Leman dimension of the incidence graph of any projective plane of order $q$ is at least $4$.

math.CO↗

arXiv2023

Practical development of efficient thermoelectric-photovoltaic hybrid systems based on wide-gap solar cells

The decrease of solar cell efficiency with temperature is a known problem for photovoltaics (PV). Temperature sensitivity can lead to a considerable amount of energy losses over the lifetime of solar panels. In this perspective Hybrid Thermoelectric-Photovoltaic (HTEPV) systems, which recover solar cell heat losses to produce an additional power output, can be a suitable option. However only hybridization of wide-gap solar cells is convenient in terms of efficiency gains and deserves investigation to evaluate HTEPV devices effectiveness. In this work we report the modeling and the development of customized bismuth telluride thermoelectric generators, optimized to be hybridized with amorphous silicon (aSi), Gallium Indium Phosphide (GaInP) or Perovskites solar cells. The model results showed in all three cases efficiency gains with a maximum of +3.1% for Perovskites (from 16.4% to 19.5%). These enhancements were then experimentally validated for the case of Perovskites solar cells, for which maximum gains were found to occur at typical operating temperatures of conventional PVs. This experimental evaluation demonstrated in an accurate fashion the real potential of thermoelectric hybridization of solar cells.

physics.app-ph↗

arXiv2023

TLM: Token-Level Masking for Transformers

Structured dropout approaches, such as attention dropout and DropHead, have been investigated to regularize the multi-head attention mechanism in Transformers. In this paper, we propose a new regularization scheme based on token-level rather than structure-level to reduce overfitting. Specifically, we devise a novel Token-Level Masking (TLM) training strategy for Transformers to regularize the connections of self-attention, which consists of two masking techniques that are effective and easy to implement. The underlying idea is to manipulate the connections between tokens in the multi-head attention via masking, where the networks are forced to exploit partial neighbors' information to produce a meaningful representation. The generality and effectiveness of TLM are thoroughly evaluated via extensive experiments on 4 diversified NLP tasks across 18 datasets, including natural language understanding benchmark GLUE, ChineseGLUE, Chinese Grammatical Error Correction, and data-to-text generation. The results indicate that TLM can consistently outperform attention dropout and DropHead, e.g., it increases by 0.5 points relative to DropHead with BERT-large on GLUE. Moreover, TLM can establish a new record on the data-to-text benchmark Rotowire (18.93 BLEU). Our code will be publicly available at https://github.com/Young1993/tlm.

cs.CL↗