arXiv subjects

Yifei Zhu

Publications and source records attributed to Yifei Zhu.

At least 55 records · Page 3Linked to original sources

arXiv2024

SpaceMeta: Global-Scale Massive Multi-User Virtual Interaction over LEO Satellite Constellations

Low latency and high synchronization among users are critical for emerging multi-user virtual interaction applications. However, the existing ground-based cloud solutions are naturally limited by the complex ground topology and fiber speeds, making it difficult to pace with the requirement of multi-user virtual interaction. The growth of low earth orbit (LEO) satellite constellations becomes a promising alternative to ground solutions. To fully exploit the potential of the LEO satellite, in this paper, we study the satellite server selection problem for global-scale multi-user interaction applications over LEO constellations. We propose an effective server selection framework, called SpaceMeta, that jointly selects the ingress satellite servers and relay servers on the communication path to minimize latency and latency discrepancy among users. Extensive experiments using real-world Starlink topology demonstrate that SpaceMeta reduces the latency by 6.72% and the interquartile range (IQR) of user latency by 39.50% compared with state-of-the-art methods.

cs.MM↗

arXiv2024

Federated Analytics-Empowered Frequent Pattern Mining for Decentralized Web 3.0 Applications

The emerging Web 3.0 paradigm aims to decentralize existing web services, enabling desirable properties such as transparency, incentives, and privacy preservation. However, current Web 3.0 applications supported by blockchain infrastructure still cannot support complex data analytics tasks in a scalable and privacy-preserving way. This paper introduces the emerging federated analytics (FA) paradigm into the realm of Web 3.0 services, enabling data to stay local while still contributing to complex web analytics tasks in a privacy-preserving way. We propose FedWeb, a tailored FA design for important frequent pattern mining tasks in Web 3.0. FedWeb remarkably reduces the number of required participating data owners to support privacy-preserving Web 3.0 data analytics based on a novel distributed differential privacy technique. The correctness of mining results is guaranteed by a theoretically rigid candidate filtering scheme based on Hoeffding's inequality and Chebychev's inequality. Two response budget saving solutions are proposed to further reduce participating data owners. Experiments on three representative Web 3.0 scenarios show that FedWeb can improve data utility by ~25.3% and reduce the participating data owners by ~98.4%.

cs.DC↗

arXiv2023

FedGKD: Unleashing the Power of Collaboration in Federated Graph Neural Networks

Federated training of Graph Neural Networks (GNN) has become popular in recent years due to its ability to perform graph-related tasks under data isolation scenarios while preserving data privacy. However, graph heterogeneity issues in federated GNN systems continue to pose challenges. Existing frameworks address the problem by representing local tasks using different statistics and relating them through a simple aggregation mechanism. However, these approaches suffer from limited efficiency from two aspects: low quality of task-relatedness quantification and inefficacy of exploiting the collaboration structure. To address these issues, we propose FedGKD, a novel federated GNN framework that utilizes a novel client-side graph dataset distillation method to extract task features that better describe task-relatedness, and introduces a novel server-side aggregation mechanism that is aware of the global collaboration structure. We conduct extensive experiments on six real-world datasets of different scales, demonstrating our framework's outperformance.

cs.LG↗

arXiv2023

Comverse: A Federative-by-Design Platform for Community Computing

Communities, ranging from homes to cities, are a ubiquitous part of our lives. However, there is a lack of adequate support for applications built around these communities. As a result, current applications each need to implement their own notion of communities, making it difficult for both the app developers and the app users (\ie the community admins and members) to create and use these community apps. In this paper, we argue that communities should be supported at the infrastructure-level rather than at the app-level. We refer to this approach as the Platform-Managed Community (PMC). We propose Comverse, a platform designed to this end. Comverse is predicated on the principle of federation, allowing autonomous nodes representing community members to voluntarily associate and share data while maintaining control over their data and participation. Through Comverse, we explore the vision of community computing by showcasing its applicability with real-world community apps.

cs.DC↗

arXiv2023

Mixed-Precision Neural Network Quantization via Learned Layer-wise Importance

The exponentially large discrete search space in mixed-precision quantization (MPQ) makes it hard to determine the optimal bit-width for each layer. Previous works usually resort to iterative search methods on the training set, which consume hundreds or even thousands of GPU-hours. In this study, we reveal that some unique learnable parameters in quantization, namely the scale factors in the quantizer, can serve as importance indicators of a layer, reflecting the contribution of that layer to the final accuracy at certain bit-widths. These importance indicators naturally perceive the numerical transformation during quantization-aware training, which can precisely provide quantization sensitivity metrics of layers. However, a deep network always contains hundreds of such indicators, and training them one by one would lead to an excessive time cost. To overcome this issue, we propose a joint training scheme that can obtain all indicators at once. It considerably speeds up the indicators training process by parallelizing the original sequential training processes. With these learned importance indicators, we formulate the MPQ search problem as a one-time integer linear programming (ILP) problem. That avoids the iterative search and significantly reduces search time without limiting the bit-width search space. For example, MPQ search on ResNet18 with our indicators takes only 0.06 s, which improves time efficiency exponentially compared to iterative search methods. Also, extensive experiments show our approach can achieve SOTA accuracy on ImageNet for far-ranging models with various constraints (e.g., BitOps, compress rate). Code is available on https://github.com/1hunters/LIMPQ.

cs.LG↗

arXiv2022

Topological classification for intersection singularities of exceptional surfaces in pseudo-Hermitian systems

Exceptional points play a pivotal role in the topology of non-Hermitian systems, and significant advances have been made in classifying exceptional points and exploring the associated phenomena. Exceptional surfaces, which are hypersurfaces of exceptional degeneracies in parameter space, can support hypersurface singularities, such as cusps, intersections and swallowtail catastrophes. Here we topologically classify the intersection singularity of exceptional surfaces for a generic pseudo-Hermitian system with parity-time symmetry. By constructing the quotient space under equivalence relations of eigenstates, we reveal that the topology of such gapless structures can be described by a non-Abelian free group on three generators. Importantly, the classification predicts a new kind of non-Hermitian gapless topological phase and can systematically explain how the exceptional surfaces and their intersections evolve under perturbations with symmetries preserved. Our work opens a new pathway for designing systems with robust topological phases, and provides inspiration for applications such as sensing and lasing which can utilize the special properties inherent in exceptional surfaces and intersections.

cond-mat.mtrl-sci↗

arXiv2022

FedWalk: Communication Efficient Federated Unsupervised Node Embedding with Differential Privacy

Node embedding aims to map nodes in the complex graph into low-dimensional representations. The real-world large-scale graphs and difficulties of labeling motivate wide studies of unsupervised node embedding problems. Nevertheless, previous effort mostly operates in a centralized setting where a complete graph is given. With the growing awareness of data privacy, data holders who are only aware of one vertex and its neighbours demand greater privacy protection. In this paper, we introduce FedWalk, a random-walk-based unsupervised node embedding algorithm that operates in such a node-level visibility graph with raw graph information remaining locally. FedWalk is designed to offer centralized competitive graph representation capability with data privacy protection and great communication efficiency. FedWalk instantiates the prevalent federated paradigm and contains three modules. We first design a hierarchical clustering tree (HCT) constructor to extract the structural feature of each node. A dynamic time warping algorithm seamlessly handles the structural heterogeneity across different nodes. Based on the constructed HCT, we then design a random walk generator, wherein a sequence encoder is designed to preserve privacy and a two-hop neighbor predictor is designed to save communication cost. The generated random walks are then used to update node embedding based on a SkipGram model. Extensive experiments on two large graphs demonstrate that Fed-Walk achieves competitive representativeness as a centralized node embedding algorithm does with only up to 1.8% Micro-F1 score and 4.4% Marco-F1 score loss while reducing about 6.7 times of inter-device communication per walk.

cs.DC↗

arXiv2022

Cost Effective MLaaS Federation: A Combinatorial Reinforcement Learning Approach

With the advancement of deep learning techniques, major cloud providers and niche machine learning service providers start to offer their cloud-based machine learning tools, also known as machine learning as a service (MLaaS), to the public. According to our measurement, for the same task, these MLaaSes from different providers have varying performance due to the proprietary datasets, models, etc. Federating different MLaaSes together allows us to improve the analytic performance further. However, naively aggregating results from different MLaaSes not only incurs significant momentary cost but also may lead to sub-optimal performance gain due to the introduction of possible false-positive results. In this paper, we propose Armol, a framework to federate the right selection of MLaaS providers to achieve the best possible analytic performance. We first design a word grouping algorithm to unify the output labels across different providers. We then present a deep combinatorial reinforcement learning based-approach to maximize the accuracy while minimizing the cost. The predictions from the selected providers are then aggregated together using carefully chosen ensemble strategies. The real-world trace-driven evaluation further demonstrates that Armol is able to achieve the same accuracy results with $67\%$ less inference cost.

cs.LG↗

arXiv2022

Arbitrary Bit-width Network: A Joint Layer-Wise Quantization and Adaptive Inference Approach

Conventional model quantization methods use a fixed quantization scheme to different data samples, which ignores the inherent "recognition difficulty" differences between various samples. We propose to feed different data samples with varying quantization schemes to achieve a data-dependent dynamic inference, at a fine-grained layer level. However, enabling this adaptive inference with changeable layer-wise quantization schemes is challenging because the combination of bit-widths and layers is growing exponentially, making it extremely difficult to train a single model in such a vast searching space and use it in practice. To solve this problem, we present the Arbitrary Bit-width Network (ABN), where the bit-widths of a single deep network can change at runtime for different data samples, with a layer-wise granularity. Specifically, first we build a weight-shared layer-wise quantizable "super-network" in which each layer can be allocated with multiple bit-widths and thus quantized differently on demand. The super-network provides a considerably large number of combinations of bit-widths and layers, each of which can be used during inference without retraining or storing myriad models. Second, based on the well-trained super-network, each layer's runtime bit-width selection decision is modeled as a Markov Decision Process (MDP) and solved by an adaptive inference strategy accordingly. Experiments show that the super-network can be built without accuracy degradation, and the bit-widths allocation of each layer can be adjusted to deal with various inputs on the fly. On ImageNet classification, we achieve 1.1% top1 accuracy improvement while saving 36.2% BitOps.

cs.CV↗

arXiv2022

Parallel Spatio-Temporal Attention-Based TCN for Multivariate Time Series Prediction

As industrial systems become more complex and monitoring sensors for everything from surveillance to our health become more ubiquitous, multivariate time series prediction is taking an important place in the smooth-running of our society. A recurrent neural network with attention to help extend the prediction windows is the current-state-of-the-art for this task. However, we argue that their vanishing gradients, short memories, and serial architecture make RNNs fundamentally unsuited to long-horizon forecasting with complex data. Temporal convolutional networks (TCNs) do not suffer from gradient problems and they support parallel calculations, making them a more appropriate choice. Additionally, they have longer memories than RNNs, albeit with some instability and efficiency problems. Hence, we propose a framework, called PSTA-TCN, that combines a parallel spatio-temporal attention mechanism to extract dynamic internal correlations with stacked TCN backbones to extract features from different window sizes. The framework makes full use parallel calculations to dramatically reduce training times, while substantially increasing accuracy with stable prediction windows up to 13 times longer than the status quo.

cs.LG↗

arXiv2020

PHOTOPiC: Calculate photo-ionization functions and model coefficients for gas discharge simulations

A program to compute photo-ionization functions and fitting parameters for an efficient photo-ionization model is presented. The code integrates the product of spectrum emission intensity, the photo-ionization yield and the absorption coefficient to calculate the photo-ionization function of each gas and the total photo-ionization function of the mixture. The coefficients of Helmholtz photo-ionization model is obtained by fitting the total photo-ionization function. A database consisting $\rm N_2$, $\rm O_2$, $\rm CO_2$ and $\rm H_2O$ molecules are included and can be modified by the users. The program provides more accurate photo-ionization functions and source terms for plasma fluid models.

physics.plasm-ph↗

arXiv2020

Modeling and theoretical analysis of SDBD plasma actuators driven by Fast-Rise-Slow-Decay Pulsed-DC voltages

Surface dielectric barrier discharge (SDBD) actuators driven by the Pulsed-DC voltages are analyzed. The Pulsed-DC SDBD studied in this work is equivalent to a classical SDBD driven by a tailored Fast-Rise-Slow-Decay (FRSD) voltage waveform. The plasma channel formation and charge production process in the voltage rising stage are studied at different slopes using a classical 2D fluid model, the thrust generated in the voltage decaying stage is studied based on an analytical approach taking 2D model results as the input. A thrust pulse is generated in the trailing edge of the voltage waveform and reaches maximum when the voltage decreases by approximately the value of cathode voltage fall~($\approx$600~V). The time duration of the rising and trailing edge, the decay rate and the amplitude of applied voltage are the main factors affecting the performance of the actuator. Analytical expressions are formulated for the value and time moment of peak thrust, the upper limit of thrust is also estimated. Higher voltage rising rate leads to higher charge density in the voltage rising stage thus higher thrust. Shorter voltage trailing edge, in general, results in higher value and earlier appearance of the peak thrust. The detailed profile of the trailing edge also affects the performance. Results in this work allow us to flexibly design the FRSD waveforms for an SDBD actuator according to the requirements of active flow control in different application conditions.

physics.plasm-ph↗

arXiv2020

The Secondary Ionization Wave and Characteristic Map of Surface Discharge Plasma in a Wide Time Scale

Variation of voltage profiles with different time scales leads to the redistribution of deposited energy as well as electro-hydrodynamic forces, while the mechanism and scaling law is still unknown. On the basis of theoretical and numerical analysis, we show that a secondary surface ionization wave forms during the voltage rising slope when electron density decreases to a critical level while the voltage is still rising. A characteristic map of energy and electro-hydrodynamics force in time scales between 1~ns and 0.1~s at atmospheric pressure is proposed, opening the door towards the target--directed design of surface discharges.

physics.plasm-ph↗

arXiv2018

Semistable models for modular curves and power operations for Morava E-theories of height 2

We construct an integral model for Lubin-Tate curves as moduli of finite subgroups of formal deformations over complete Noetherian local rings. They are p-adic completions of the modular curves X_0(p) at a mod-p supersingular point. Our model is semistable in the sense that the only singularities of its special fiber are normal crossings. Given this model, we obtain a uniform presentation for the Dyer-Lashof algebra of Morava E-theories at height 2 as local moduli of power operations in elliptic cohomology.

math.AT↗

arXiv2018

A Capsule Network for Traffic Speed Prediction in Complex Road Networks

This paper proposes a deep learning approach for traffic flow prediction in complex road networks. Traffic flow data from induction loop sensors are essentially a time series, which is also spatially related to traffic in different road segments. The spatio-temporal traffic data can be converted into an image where the traffic data are expressed in a 3D space with respect to space and time axes. Although convolutional neural networks (CNNs) have been showing surprising performance in understanding images, they have a major drawback. In the max pooling operation, CNNs are losing important information by locally taking the highest activation values. The inter-relationship in traffic data measured by sparsely located sensors in different time intervals should not be neglected in order to obtain accurate predictions. Thus, we propose a neural network with capsules that replaces max pooling by dynamic routing. This is the first approach that employs the capsule network on a time series forecasting problem, to our best knowledge. Moreover, an experiment on real traffic speed data measured in the Santander city of Spain demonstrates the proposed method outperforms the state-of-the-art method based on a CNN by 13.1% in terms of root mean squared error.

cs.CV↗

arXiv2017

Norm coherence for descent of level structures on formal deformations

We give a formulation for descent of level structures on deformations of formal groups, and study the compatibility between the descent and a norm construction. Under this framework, we generalize Ando's construction of H-infinity complex orientations for Morava E-theories associated to Honda formal group laws over F_p. We show the existence and uniqueness of such an orientation for any Morava E-theory associated to a formal group law over an algebraic extension of F_p and, in particular, orientations for a family of elliptic cohomology theories. These orientations correspond to coordinates on deformations of formal groups which are compatible with norm maps along descent.

math.AT↗

arXiv2016

Morava E-homology of Bousfield-Kuhn functors on odd-dimensional spheres

As an application of Behrens and Rezk's spectral algebra model for unstable v_n-periodic homotopy theory, we give explicit presentations for the completed E-homology of the Bousfield-Kuhn functor on odd-dimensional spheres at chromatic level 2, and compare them to the level 1 case. The latter reflects earlier work in the literature on K-theory localizations.

math.AT↗

arXiv2015

The Hecke algebra action on Morava E-theory of height 2

Given a one-dimensional formal group of height 2, let E be the Morava E-theory spectrum associated to its universal deformation over the Lubin-Tate ring. By computing with moduli spaces of elliptic curves, we give an explicitation for an algebra of Hecke operators acting on E-cohomology. This leads to a vanishing result for Rezk's logarithmic cohomology operation on the units of E. It identifies a family of elements in the kernel with meromorphic modular forms whose Serre derivative is zero. Our calculation finds a connection to logarithms of modular units. In particular, we work out an action of Hecke operators on certain "logarithmic" q-series, in the sense of Knopp and Mason, that agrees with our vanishing result and extends the classical Hecke action on modular forms.

math.AT↗