arXiv subjects

Meng Li

Publications and source records attributed to Meng Li.

At least 271 records · Page 15Linked to original sources

arXiv2023

On the Estimation of Derivatives Using Plug-in Kernel Ridge Regression Estimators

We study the problem of estimating the derivatives of a regression function, which has a wide range of applications as a key nonparametric functional of unknown functions. Standard analysis may be tailored to specific derivative orders, and parameter tuning remains a daunting challenge particularly for high-order derivatives. In this article, we propose a simple plug-in kernel ridge regression (KRR) estimator in nonparametric regression with random design that is broadly applicable for multi-dimensional support and arbitrary mixed-partial derivatives. We provide a non-asymptotic analysis to study the behavior of the proposed estimator in a unified manner that encompasses the regression function and its derivatives, leading to two error bounds for a general class of kernels under the strong $L_\infty$ norm. In a concrete example specialized to kernels with polynomially decaying eigenvalues, the proposed estimator recovers the minimax optimal rate up to a logarithmic factor for estimating derivatives of functions in Hölder and Sobolev classes. Interestingly, the proposed estimator achieves the optimal rate of convergence with the same choice of tuning parameter for any order of derivatives. Hence, the proposed estimator enjoys a \textit{plug-in property} for derivatives in that it automatically adapts to the order of derivatives to be estimated, enabling easy tuning in practice. Our simulation studies show favorable finite sample performance of the proposed method relative to several existing methods and corroborate the theoretical findings on its minimax optimality.

stat.ML↗

arXiv2023

Memory-aware Scheduling for Complex Wired Networks with Iterative Graph Optimization

Memory-aware network scheduling is becoming increasingly important for deep neural network (DNN) inference on resource-constrained devices. However, due to the complex cell-level and network-level topologies, memory-aware scheduling becomes very challenging. While previous algorithms all suffer from poor scalability, in this paper, we propose an efficient memory-aware scheduling framework based on iterative computation graph optimization. Our framework features an iterative graph fusion algorithm that simplifies the computation graph while preserving the scheduling optimality. We further propose an integer linear programming formulation together with topology-aware variable pruning to schedule the simplified graph efficiently. We evaluate our method against prior-art algorithms on different networks and demonstrate that our method outperforms existing techniques in all the benchmarks, reducing the peak memory footprint by 13.4%, and achieving better scalability for networks with complex network-level topologies.

cs.LG↗

arXiv2023

Falcon: Accelerating Homomorphically Encrypted Convolutions for Efficient Private Mobile Network Inference

Efficient networks, e.g., MobileNetV2, EfficientNet, etc, achieves state-of-the-art (SOTA) accuracy with lightweight computation. However, existing homomorphic encryption (HE)-based two-party computation (2PC) frameworks are not optimized for these networks and suffer from a high inference overhead. We observe the inefficiency mainly comes from the packing algorithm, which ignores the computation characteristics and the communication bottleneck of homomorphically encrypted depthwise convolutions. Therefore, in this paper, we propose Falcon, an effective dense packing algorithm for HE-based 2PC frameworks. Falcon features a zero-aware greedy packing algorithm and a communication-aware operator tiling strategy to improve the packing density for depthwise convolutions. Compared to SOTA HE-based 2PC frameworks, e.g., CrypTFlow2, Iron and Cheetah, Falcon achieves more than 15.6x, 5.1x and 1.8x latency reduction, respectively, at operator level. Meanwhile, at network level, Falcon allows for 1.4% and 4.2% accuracy improvement over Cheetah on CIFAR-100 and TinyImagenet datasets with iso-communication, respecitvely.

cs.CR↗

arXiv2023

Atomic scale understanding of initial Cu-Ni oxidation from machine-learning accelerated first-principles simulations and in situ TEM experiments

The development of accurate methods for determining how alloy surfaces spontaneously restructure under reactive and corrosive environments is a key, long-standing, grand challenge in materials science. Current oxidation models, such as Cabrera-Mott, are based on macroscopic empirical knowledge that lacks fundamental insight at the atomic level. Using machine learning-accelerated density functional theory with in situ environmental transmission electron microscopy (ETEM), we examine the interplay between surface reconstructions and preferential segregation tendencies of CuNi(100) surfaces under oxidation conditions. Our modeling approach based on molecular dynamics and grand canonical Monte Carlo simulations shows that oxygen-induced Ni segregation in CuNi alloy favors Cu(100)-O c(2x2) reconstruction and destabilizes the Cu(100)-O missing row reconstruction. The underpinnings of these stabilization tendencies are rationalized based on the similar atomic coordination and bond lengths in NiO rock salt and Cu(100)-O c(2x2) structures. In situ ETEM experiments show Ni segregation followed by NiO nucleation and growth in regions without MRR, with secondary nucleation and growth of Cu2O in MRR regions. This further corroborates the simulated surface oxidation and segregation modelling outcomes. Our findings are general and are expected to extend to other alloy systems.

cond-mat.mtrl-sci↗

arXiv2023

MPCViT: Searching for Accurate and Efficient MPC-Friendly Vision Transformer with Heterogeneous Attention

Secure multi-party computation (MPC) enables computation directly on encrypted data and protects both data and model privacy in deep learning inference. However, existing neural network architectures, including Vision Transformers (ViTs), are not designed or optimized for MPC and incur significant latency overhead. We observe Softmax accounts for the major latency bottleneck due to a high communication complexity, but can be selectively replaced or linearized without compromising the model accuracy. Hence, in this paper, we propose an MPC-friendly ViT, dubbed MPCViT, to enable accurate yet efficient ViT inference in MPC. Based on a systematic latency and accuracy evaluation of the Softmax attention and other attention variants, we propose a heterogeneous attention optimization space. We also develop a simple yet effective MPC-aware neural architecture search algorithm for fast Pareto optimization. To further boost the inference efficiency, we propose MPCViT+, to jointly optimize the Softmax attention and other network components, including GeLU, matrix multiplication, etc. With extensive experiments, we demonstrate that MPCViT achieves 1.9%, 1.3% and 3.6% higher accuracy with 6.2x, 2.9x and 1.9x latency reduction compared with baseline ViT, MPCFormer and THE-X on the Tiny-ImageNet dataset, respectively. MPCViT+ further achieves a better Pareto front compared with MPCViT. The code and models for evaluation are available at https://github.com/PKU-SEC-Lab/mpcvit.

cs.CR↗

arXiv2023

Decentralized Threshold Signatures with Dynamically Private Accountability

Threshold signatures are a fundamental cryptographic primitive used in many practical applications. As proposed by Boneh and Komlo (CRYPTO'22), TAPS is a threshold signature that is a hybrid of privacy and accountability. It enables a combiner to combine t signature shares while revealing nothing about the threshold t or signing quorum to the public and asks a tracer to track a signature to the quorum that generates it. However, TAPS has three disadvantages: it 1) structures upon a centralized model, 2) assumes that both combiner and tracer are honest, and 3) leaves the tracing unnotarized and static. In this work, we introduce Decentralized, Threshold, dynamically Accountable and Private Signature (DeTAPS) that provides decentralized combining and tracing, enhanced privacy against untrusted combiners (tracers), and notarized and dynamic tracing. Specifically, we adopt Dynamic Threshold Public-Key Encryption (DTPKE) to dynamically notarize the tracing process, design non-interactive zero knowledge proofs to achieve public verifiability of notaries, and utilize the Key-Aggregate Searchable Encryption to bridge TAPS and DTPKE so as to awaken the notaries securely and efficiently. In addition, we formalize the definitions and security requirements for DeTAPS. Then we present a generic construction and formally prove its security and privacy. To evaluate the performance, we build a prototype based on SGX2 and Ethereum.

cs.CR↗

arXiv2023

Topological Insulator VxBi1.08-xSn0.02Sb0.9Te2S as a Promising n-type Thermoelectric Material

As one of the most important n-type thermoelectric (TE) materials, Bi2Te3 has been studied for decades, with efforts to enhance the thermoelectric performance based on element doping, band engineering, etc. In this study, we report a novel bulk-insulating topological material system as a replacement for n-type Bi2Te3 materials: V doped Bi1.08Sn0.02Sb0.9Te2S (V:BSSTS) . The V:BSSTS is a bulk insulator with robust metallic topological surface states. Furthermore, the bulk band gap can be tuned by the doping level of V, which is verified by magnetotransport measurements. Large linear magnetoresistance is observed in all samples. Excellent thermoelectric performance is obtained in the V:BSSTS samples, e.g., the highest figure of merit ZT of ~ 0.8 is achieved in the 2% V doped sample (denoted as V0.02) at 550 K. The high thermoelectric performance of V:BSSTS can be attributed to two synergistic effects: (1) the low conductive secondary phases Sb2S3, and V2S3 are believed to be important scattering centers for phonons, leading to lower lattice thermal conductivity; and (2) the electrical conductivity is increased due to the high-mobility topological surface states at the boundaries. In addition, by replacing one third of costly tellurium with abundant, low-cost, and less-toxic sulfur element, the newly produced BSSTS material is inexpensive but still has comparable TE performance to the traditional Bi2Te3-based materials, which offers a cheaper plan for the electronics and thermoelectric industries. Our results demonstrate that topological materials with unique band structures can provide a new platform in the search for new high performance TE materials.

cond-mat.mtrl-sci↗

arXiv2023

Study of (3He, t) charge exchange reactions to isobaric analog states in inverse kinematics

The transition between isobaric analog states (IAS) in the (3He, t) charge exchange reaction presents a unique opportunity to access the isospin structure of the nuclei. In this study not only the Fermi transition but also the Gamow-Teller (G-T) transition of the IAS reaction were investigated for the 13,14C(3He, t) and 17,18,19,20O(3He, t) reactions, in order to explore the neutron number dependence of the IAS reaction for the light neutron-rich nuclei. It was found that the G-T type IAS reaction also exhibited a significant dependence of the transition strength on the neutron number and the angular momentum configuration of the nuclei. Additionally, the inverse kinematics was also discussed for extracting the yields of the interested reaction channels in the proposed experiments on radioactive beams. The calculated triton yields demonstrated the capability of the proposed experiments to obtain meaningful results.

nucl-th↗

arXiv2023

Ultra Electron Density Sensitivity for Surface Plasmons

We investigate surface plasmons from a solid-state standpoint and highlight their ultra electron density sensitivity. When a surface plasmon is excited on a planar gold film by an evanescent wave from 625 nm incident light, only a minute fraction of the surface electron density, approximately one thousandth, participates in the process. By introducing a noise-depressed surface potential modulation, we reduce the electron density to the order of 10 um-2, enabling electron sensitivity on the order of 0.1 e. As a practical application, we develop a surface plasmon resonance imaging method capable of detecting single anions in solution at a concentration of 1 aM.

physics.optics↗

arXiv2023

Stroke Extraction of Chinese Character Based on Deep Structure Deformable Image Registration

Stroke extraction of Chinese characters plays an important role in the field of character recognition and generation. The most existing character stroke extraction methods focus on image morphological features. These methods usually lead to errors of cross strokes extraction and stroke matching due to rarely using stroke semantics and prior information. In this paper, we propose a deep learning-based character stroke extraction method that takes semantic features and prior information of strokes into consideration. This method consists of three parts: image registration-based stroke registration that establishes the rough registration of the reference strokes and the target as prior information; image semantic segmentation-based stroke segmentation that preliminarily separates target strokes into seven categories; and high-precision extraction of single strokes. In the stroke registration, we propose a structure deformable image registration network to achieve structure-deformable transformation while maintaining the stable morphology of single strokes for character images with complex structures. In order to verify the effectiveness of the method, we construct two datasets respectively for calligraphy characters and regular handwriting characters. The experimental results show that our method strongly outperforms the baselines. Code is available at https://github.com/MengLi-l1/StrokeExtraction.

cs.CV↗

arXiv2023

Frame-Event Alignment and Fusion Network for High Frame Rate Tracking

Most existing RGB-based trackers target low frame rate benchmarks of around 30 frames per second. This setting restricts the tracker's functionality in the real world, especially for fast motion. Event-based cameras as bioinspired sensors provide considerable potential for high frame rate tracking due to their high temporal resolution. However, event-based cameras cannot offer fine-grained texture information like conventional cameras. This unique complementarity motivates us to combine conventional frames and events for high frame rate object tracking under various challenging conditions. Inthispaper, we propose an end-to-end network consisting of multi-modality alignment and fusion modules to effectively combine meaningful information from both modalities at different measurement rates. The alignment module is responsible for cross-style and cross-frame-rate alignment between frame and event modalities under the guidance of the moving cues furnished by events. While the fusion module is accountable for emphasizing valuable features and suppressing noise information by the mutual complement between the two modalities. Extensive experiments show that the proposed approach outperforms state-of-the-art trackers by a significant margin in high frame rate tracking. With the FE240hz dataset, our approach achieves high frame rate tracking up to 240Hz.

cs.CV↗

arXiv2023

On the Softplus Penalty for Constrained Convex Optimization

We study a new penalty reformulation of constrained convex optimization based on the softplus penalty function. We develop novel and tight upper bounds on the objective value gap and the violation of constraints for the solutions to the penalty reformulations by analyzing the solution path of the reformulation with respect to the smoothness parameter. We use these upper bounds to analyze the complexity of applying gradient methods, which are advantageous when the number of constraints is large, to the reformulation.

math.OC↗

arXiv2023

Direct imaging of single-molecule electronic states in chemical reactions

Current single-molecule imaging techniques are incapable of providing stochastic information of individual chemical reactions. We have developed an optical imaging method for stochastic analysis of individual electrochemical reactions. The method bypasses diffraction limits and allows for the direct imaging of the electronic states of single molecules in electrochemical reactions. It involves using surface plasmon resonance and electrode potential modulation to reduce electron density background on macroelectrode surfaces, thus making it possible to measure individual redox reactions, identify different redox states of single molecules, and analyze reaction probability fluctuations under different conditions. The technique enables the acquisition of electron transfer and stochastic information in chemical processes, providing unique insights into single-molecule reactions with broad applications in biology, materials, computational science, and other scientific communities.

physics.optics↗

arXiv2023

ALT: An Automatic System for Long Tail Scenario Modeling

In this paper, we consider the problem of long tail scenario modeling with budget limitation, i.e., insufficient human resources for model training stage and limited time and computing resources for model inference stage. This problem is widely encountered in various applications, yet has received deficient attention so far. We present an automatic system named ALT to deal with this problem. Several efforts are taken to improve the algorithms used in our system, such as employing various automatic machine learning related techniques, adopting the meta learning philosophy, and proposing an essential budget-limited neural architecture search method, etc. Moreover, to build the system, many optimizations are performed from a systematic perspective, and essential modules are armed, making the system more feasible and efficient. We perform abundant experiments to validate the effectiveness of our system and demonstrate the usefulness of the critical modules in our system. Moreover, online results are provided, which fully verified the efficacy of our system.

cs.LG↗

arXiv2023

Zero-Knowledge Proof-based Practical Federated Learning on Blockchain

Since the concern of privacy leakage extremely discourages user participation in sharing data, federated learning has gradually become a promising technique for both academia and industry for achieving collaborative learning without leaking information about the local data. Unfortunately, most federated learning solutions cannot efficiently verify the execution of each participant's local machine learning model and protect the privacy of user data, simultaneously. In this article, we first propose a Zero-Knowledge Proof-based Federated Learning (ZKP-FL) scheme on blockchain. It leverages zero-knowledge proof for both the computation of local data and the aggregation of local model parameters, aiming to verify the computation process without requiring the plaintext of the local data. We further propose a Practical ZKP-FL (PZKP-FL) scheme to support fraction and non-linear operations. Specifically, we explore a Fraction-Integer mapping function, and use Taylor expansion to efficiently handle non-linear operations while maintaining the accuracy of the federated learning model. We also analyze the security of PZKP-FL. Performance analysis demonstrates that the whole running time of the PZKP-FL scheme is approximately less than one minute in parallel execution.

cs.CR↗

arXiv2023

Development of an eReaxFF Force Field for BZY20 Solid Oxide Electrocatalysis

Electrocatalysis is a catalytic process where the rate of an electrochemical reaction occurring at the electrode-electrolyte interface can be controlled by varying the electrical potential. Electrocatalysis can be applied to generate hydrogen which can be stored for future use in fuel cells for clean electricity. The use of solid oxide in electrocatalysis specially in hydrogen evolution reaction is promising. However, further improvements are essential in order to meet the ever-increasing global energy demand. Improvement of the performance of these high energy chemical systems is directly linked to the understanding and improving the complex physical and chemical phenomena and exchanges that take place at their different interfaces. To enable large length and time scale atomistic simulations of solid oxide electrocatalysis for hydrogen generation, we developed an eReaxFF force field for barium zirconate doped with 20 mol% of yttrium (BZY20). All parameters for the eReaxFF were optimized to reproduce quantum mechanical (QM) calculations on relevant condensed phase and cluster systems describing oxygen vacancies, vacancy migrations, water adsorption, water splitting and hydrogen generation on the surfaces of the BZY20 solid oxide. Using the developed force field, we performed zero-voltage molecular dynamics simulations to observe water adsorption and the eventual hydrogen production. Based on our simulation results, we conclude that this force field sets a stage for the introduction of explicit electron concept in order to simulate electron conductivity, electron leakage and non-zero-voltage effects on hydrogen generation. Overall, we demonstrate how atomistic-scale simulations can enhance our understanding of processes at interfaces in solid oxide materials.

physics.chem-ph↗

arXiv2023

Eigenproblems in addition-min algebra

In order to guarantee the downloading quality requirements of users and improve the stability of data transmission in a BitTorrent-like peer-to-peer file sharing system, this article deals with eigenproblems of addition-min algebras. First, it provides a sufficient and necessary condition for a vector being an eigenvector of a given matrix, and then presents an algorithm for finding all eigenvalues and eigenvectors of a given matrix. It further proposes a sufficient and necessary condition for a vector being a constrained eigenvector of a given matrix and supplies an algorithm for computing all the constrained eigenvectors and eigenvalues of a given matrix. This article finally discusses the supereigenproblem of a given matrix and presents an algorithm for obtaining the maximum constrained supereigenvalue and depicting the feasible region of all the constrained supereigenvectors for a given matrix. It also gives some examples for illustrating the algorithms, respectively.

math.GM↗

arXiv2023

${S}^{2}$Net: Accurate Panorama Depth Estimation on Spherical Surface

Monocular depth estimation is an ambiguous problem, thus global structural cues play an important role in current data-driven single-view depth estimation methods. Panorama images capture the complete spatial information of their surroundings utilizing the equirectangular projection which introduces large distortion. This requires the depth estimation method to be able to handle the distortion and extract global context information from the image. In this paper, we propose an end-to-end deep network for monocular panorama depth estimation on a unit spherical surface. Specifically, we project the feature maps extracted from equirectangular images onto unit spherical surface sampled by uniformly distributed grids, where the decoder network can aggregate the information from the distortion-reduced feature maps. Meanwhile, we propose a global cross-attention-based fusion module to fuse the feature maps from skip connection and enhance the ability to obtain global context. Experiments are conducted on five panorama depth estimation datasets, and the results demonstrate that the proposed method substantially outperforms previous state-of-the-art methods. All related codes will be open-sourced in the upcoming days.

cs.CV↗