arXiv subjects

Yong Zhang

Publications and source records attributed to Yong Zhang.

At least 397 records · Page 22Linked to original sources

arXiv2022

Augmenting Operations Research with Auto-Formulation of Optimization Models from Problem Descriptions

We describe an augmented intelligence system for simplifying and enhancing the modeling experience for operations research. Using this system, the user receives a suggested formulation of an optimization problem based on its description. To facilitate this process, we build an intuitive user interface system that enables the users to validate and edit the suggestions. We investigate controlled generation techniques to obtain an automatic suggestion of formulation. Then, we evaluate their effectiveness with a newly created dataset of linear programming problems drawn from various application domains.

cs.CL↗

arXiv2022

Unsupervised Sentence Textual Similarity with Compositional Phrase Semantics

Measuring Sentence Textual Similarity (STS) is a classic task that can be applied to many downstream NLP applications such as text generation and retrieval. In this paper, we focus on unsupervised STS that works on various domains but only requires minimal data and computational resources. Theoretically, we propose a light-weighted Expectation-Correction (EC) formulation for STS computation. EC formulation unifies unsupervised STS approaches including the cosine similarity of Additively Composed (AC) sentence embeddings, Optimal Transport (OT), and Tree Kernels (TK). Moreover, we propose the Recursive Optimal Transport Similarity (ROTS) algorithm to capture the compositional phrase semantics by composing multiple recursive EC formulations. ROTS finishes in linear time and is faster than its predecessors. ROTS is empirically more effective and scalable than previous approaches. Extensive experiments on 29 STS tasks under various settings show the clear advantage of ROTS over existing approaches. Detailed ablation studies demonstrate the effectiveness of our approaches.

cs.CL↗

arXiv2022

Routing and QoS Policy Optimization in SD-WAN

In modern SD-WAN networks, a global controller continuously optimizes application and user intents by selecting the proper routing policies for each application. Nevertheless, the competition between flows can still occur at each overlay link and it may degrade Quality of Service (QoS). To mitigate performance degradations in case of congestion, QoS policies can also be dynamically optimized to limit the rate of low-priority flows and share the bandwidth among all flows in a fair manner. This demonstration presents a comprehensive control plane architecture to coordinate path selection and rate allocation, in order to meet application requirements while optimizing a global objective (e.g., low congestion, best quality, and minimum cost).

cs.NI↗

arXiv2022

On optimal zero-padding of kernel truncation method

The kernel truncation method (KTM) is a commonly-used algorithm to compute the convolution-type nonlocal potential $Φ(x)=(U\ast ρ)(x), ~x \in {\mathbb R^d}$, where the convolution kernel $U(x)$ might be singular at the origin and/or far-field and the density $ρ(x)$ is smooth and fast-decaying. In KTM, in order to capture the Fourier integrand's oscillations that is brought by the kernel truncation, one needs to carry out a zero-padding of the density, which means a larger physical computation domain and a finer mesh in the Fourier space by duality. The empirical fourfold zero-padding [ Vico et al J. Comput. Phys. (2016) ] puts a heavy burden on memory requirement especially for higher dimension problems. In this paper, we derive the optimal zero-padding factor, that is, $\sqrt{d}+1$, for the first time together with a rigorous proof. The memory cost is greatly reduced to a small fraction, i.e., $(\frac{\sqrt{d}+1}{4})^d$, of what is needed in the original fourfold algorithm. For example, in the precomputation step, a double-precision computation on a $256^3$ grid requires a minimum $3.4$ Gb memory with the optimal threefold zero-padding, while the fourfold algorithm requires around $8$ Gb where the reduction factor is $\frac{37}{64}\approx \frac{3}{5}$. Then, we present the error estimates of the potential and density in $d$ dimension. Next, we re-investigate the optimal zero-padding factor for the anisotropic density. Finally, extensive numerical results are provided to confirm the accuracy, efficiency, optimal zero-padding factor for the anisotropic density, together with some applications to different types of nonlocal potential, including the 1D/2D/3D Poisson, 2D Coulomb, quasi-2D/3D Dipole-Dipole Interaction and 3D quadrupolar potential.

math.NA↗

arXiv2022

A multi view multi stage and multi window framework for pulmonary artery segmentation from CT scans

This is the technical report of the 9th place in the final result of PARSE2022 Challenge. We solve the segmentation problem of the pulmonary artery by using a two-stage method based on a 3D CNN network. The coarse model is used to locate the ROI, and the fine model is used to refine the segmentation result. In addition, in order to improve the segmentation performance, we adopt multi-view and multi-window level method, at the same time we employ a fine-tune strategy to mitigate the impact of inconsistent labeling.

eess.IV↗

arXiv2022

Estimating Visual Information From Audio Through Manifold Learning

We propose a new framework for extracting visual information about a scene only using audio signals. Audio-based methods can overcome some of the limitations of vision-based methods i.e., they do not require "line-of-sight", are robust to occlusions and changes in illumination, and can function as a backup in case vision/lidar sensors fail. Therefore, audio-based methods can be useful even for applications in which only visual information is of interest Our framework is based on Manifold Learning and consists of two steps. First, we train a Vector-Quantized Variational Auto-Encoder to learn the data manifold of the particular visual modality we are interested in. Second, we train an Audio Transformation network to map multi-channel audio signals to the latent representation of the corresponding visual sample. We show that our method is able to produce meaningful images from audio using a publicly available audio/visual dataset. In particular, we consider the prediction of the following visual modalities from audio: depth and semantic segmentation. We hope the findings of our work can facilitate further research in visual information extraction from audio. Code is available at: https://github.com/ubc-vision/audio_manifold.

cs.CV↗

arXiv2022

Achieving Model Fairness in Vertical Federated Learning

Vertical federated learning (VFL) has attracted greater and greater interest since it enables multiple parties possessing non-overlapping features to strengthen their machine learning models without disclosing their private data and model parameters. Similar to other machine learning algorithms, VFL faces demands and challenges of fairness, i.e., the learned model may be unfairly discriminatory over some groups with sensitive attributes. To tackle this problem, we propose a fair VFL framework in this work. First, we systematically formulate the problem of training fair models in VFL, where the learning task is modelled as a constrained optimization problem. To solve it in a federated and privacy-preserving manner, we consider the equivalent dual form of the problem and develop an asynchronous gradient coordinate-descent ascent algorithm, where some active data parties perform multiple parallelized local updates per communication round to effectively reduce the number of communication rounds. The messages that the server sends to passive parties are deliberately designed such that the information necessary for local updates is released without intruding on the privacy of data and sensitive attributes. We rigorously study the convergence of the algorithm when applied to general nonconvex-concave min-max problems. We prove that the algorithm finds a $δ$-stationary point of the dual objective in $\mathcal{O}(δ^{-4})$ communication rounds under mild conditions. Finally, the extensive experiments on three benchmark datasets demonstrate the superior performance of our method in training fair models.

cs.LG↗

arXiv2022

A deep learning-based remaining useful life prediction approach for bearings

In industrial applications, nearly half the failures of motors are caused by the degradation of rolling element bearings (REBs). Therefore, accurately estimating the remaining useful life (RUL) for REBs are of crucial importance to ensure the reliability and safety of mechanical systems. To tackle this challenge, model-based approaches are often limited by the complexity of mathematical modeling. Conventional data-driven approaches, on the other hand, require massive efforts to extract the degradation features and construct health index. In this paper, a novel online data-driven framework is proposed to exploit the adoption of deep convolutional neural networks (CNN) in predicting the RUL of bearings. More concretely, the raw vibrations of training bearings are first processed using the Hilbert-Huang transform (HHT) and a novel nonlinear degradation indicator is constructed as the label for learning. The CNN is then employed to identify the hidden pattern between the extracted degradation indicator and the vibration of training bearings, which makes it possible to estimate the degradation of the test bearings automatically. Finally, testing bearings' RULs are predicted by using a $ε$-support vector regression model. The superior performance of the proposed RUL estimation framework, compared with the state-of-the-art approaches, is demonstrated through the experimental results. The generality of the proposed CNN model is also validated by transferring to bearings undergoing different operating conditions.

cs.LG↗

arXiv2022

Towards Real-World Video Deblurring by Exploring Blur Formation Process

This paper aims at exploring how to synthesize close-to-real blurs that existing video deblurring models trained on them can generalize well to real-world blurry videos. In recent years, deep learning-based approaches have achieved promising success on video deblurring task. However, the models trained on existing synthetic datasets still suffer from generalization problems over real-world blurry scenarios with undesired artifacts. The factors accounting for the failure remain unknown. Therefore, we revisit the classical blur synthesis pipeline and figure out the possible reasons, including shooting parameters, blur formation space, and image signal processor~(ISP). To analyze the effects of these potential factors, we first collect an ultra-high frame-rate (940 FPS) RAW video dataset as the data basis to synthesize various kinds of blurs. Then we propose a novel realistic blur synthesis pipeline termed as RAW-Blur by leveraging blur formation cues. Through numerous experiments, we demonstrate that synthesizing blurs in the RAW space and adopting the same ISP as the real-world testing data can effectively eliminate the negative effects of synthetic data. Furthermore, the shooting parameters of the synthesized blurry video, e.g., exposure time and frame-rate play significant roles in improving the performance of deblurring models. Impressively, the models trained on the blurry data synthesized by the proposed RAW-Blur pipeline can obtain more than 5dB PSNR gain against those trained on the existing synthetic blur datasets. We believe the novel realistic synthesis pipeline and the corresponding RAW video dataset can help the community to easily construct customized blur datasets to improve real-world video deblurring performance largely, instead of laboriously collecting real data pairs.

cs.CV↗

arXiv2022

Magnetic field control of the near-field radiative heat transfer in three-body planar systems

Recently, the application of an external magnetic field to actively control the near-field heat transfer has emerged as an appealing and promising technique. Existing studies have shown that an external static magnetic field tends to reduce the subwavelength radiative flux exchanged between two planar structures containing magneto-optical (MO) materials, but so far the nearfield thermomagnetic effects in systems with more such structures at different temperatures have not been reported. Here, we are focused on examining how the presence of an external magnetic field modifies the radiative energy transfer in a many-body configuration consisting of three MO n-doped semiconductors slabs, separated by subwavelength vacuum gaps. To exactly calculate the radiative flux transferred in such an anisotropic planar system, a general Green-function-based approach is offered, which allows one to investigate the radiative heat transfer in arbitrary manybody systems with planar geometry. We demonstrate that, under specific choices of the geometrical and thermal parameters, the applied magnetic field is able to either reduce or enhance the near-field energy transfer in three-element MO planar systems, depending on the interplay between the damped evanescent fields of the zero-field surface waves and the propagating hyperbolic modes induced by magnetic fields. Our study broadens the understanding concerning to the use of external fields to actively control the heat transfer in subwavelength regimes, and may be leveraged for potential applications in the realm of nanoscale thermal management.

physics.optics↗

arXiv2022

One-loop diagrams with quadratic propagators from the worldsheet

It is well known that forward limits of tree-level amplitudes (and those trivalent diagrams they consist of) produce one-loop amplitudes and trivalent diagrams with propagators linear in the loop momentum. They naturally arise from one-loop worldsheet formulae, and an important open problem is how to recombine them into usual one-loop diagrams with quadratic propagators. In this paper, we study a new collection of worldsheet functions: generalized one-loop Parke-Taylor factors with tensor numerators, which are conjectured to serve as a basis for one-loop worldsheet functions with this nice property. We present all-multiplicity, closed-form expressions for combinations of one-loop trivalent diagrams with quadratic propagators and tensor numerators to arbitrary rank (including possible tadpole contributions), produced by any pair of Parke-Taylor factors. We also briefly comment on reducing worldsheet functions onto such a basis, and applications to one-loop amplitudes in physical theories.

hep-th↗

arXiv2022

ATOMS: ALMA Three-millimeter Observations of Massive Star-forming regions -- XII: Fragmentation and multi-scale gas kinematics in protoclusters G12.42+0.50 and G19.88-0.53

We present new continuum and molecular line data from the ALMA Three-millimeter Observations of Massive Star-forming regions (ATOMS) survey for the two protoclusters, G12.42+0.50 and G19.88-0.53. The 3 mm continuum maps reveal seven cores in each of the two globally contracting protoclusters. These cores satisfy the radius-mass relation and the surface mass density criteria for high-mass star formation. Similar to their natal clumps, the virial analysis of the cores suggests that they are undergoing gravitational collapse ($\rm α_{vir} << 2$). The clump to core scale fragmentation is investigated and the derived core masses and separations are found to be consistent with thermal Jeans fragmentation. We detect large-scale filamentary structures with velocity gradients and multiple outflows in both regions. Dendrogram analysis of the H$^{13}$CO$^{+}$ map identifies several branch and leaf structures with sizes $\sim$ 0.1 and 0.03 pc, respectively. The supersonic gas motion displayed by the branch structures is in agreement with the Larson power-law indicating that the gas kinematics at this spatial scale is driven by turbulence. The transition to transonic/subsonic gas motion is seen to occur at spatial scales of $\sim$0.1 pc indicating the dissipation of turbulence. In agreement with this, the leaf structures reveal gas motions that deviate from the slope of Larson's law. From the large-scale converging filaments to the collapsing cores, the gas dynamics in G12.42+0.50 and G19.88-0.53 show scale-dependent dominance of turbulence and gravity and the combination of these two driving mechanisms needs to be invoked to explain massive star formation in the protoclusters.

astro-ph.GA↗

arXiv2022

Knowledge-Injected Federated Learning

Federated learning is an emerging technique for training models from decentralized data sets. In many applications, data owners participating in the federated learning system hold not only the data but also a set of domain knowledge. Such knowledge includes human know-how and craftsmanship that can be extremely helpful to the federated learning task. In this work, we propose a federated learning framework that allows the injection of participants' domain knowledge, where the key idea is to refine the global model with knowledge locally. The scenario we consider is motivated by a real industry-level application, and we demonstrate the effectiveness of our approach to this application.

cs.LG↗

arXiv2022

SemAug: Semantically Meaningful Image Augmentations for Object Detection Through Language Grounding

Data augmentation is an essential technique in improving the generalization of deep neural networks. The majority of existing image-domain augmentations either rely on geometric and structural transformations, or apply different kinds of photometric distortions. In this paper, we propose an effective technique for image augmentation by injecting contextually meaningful knowledge into the scenes. Our method of semantically meaningful image augmentation for object detection via language grounding, SemAug, starts by calculating semantically appropriate new objects that can be placed into relevant locations in the image (the what and where problems). Then it embeds these objects into their relevant target locations, thereby promoting diversity of object instance distribution. Our method allows for introducing new object instances and categories that may not even exist in the training set. Furthermore, it does not require the additional overhead of training a context network, so it can be easily added to existing architectures. Our comprehensive set of evaluations showed that the proposed method is very effective in improving the generalization, while the overhead is negligible. In particular, for a wide range of model architectures, our method achieved ~2-4% and ~1-2% mAP improvements for the task of object detection on the Pascal VOC and COCO datasets, respectively.

cs.CV↗

arXiv2022

How and what to learn:The modes of machine learning

Despite their great success, neural networks still remain as black-boxes due to the lack of interpretability. Here we propose a new analyzing method, namely the weight pathway analysis (WPA), to make them transparent. We consider weights in pathways that link neurons longitudinally from input neurons to output neurons, or simply weight pathways, as the basic units for understanding a neural network, and decompose a neural network into a series of subnetworks of such weight pathways. A visualization scheme of the subnetworks is presented that gives longitudinal perspectives of the network like radiographs, making the internal structures of the network visible. Impacts of parameter adjustments or structural changes to the network can be visualized via such radiographs. Characteristic maps are established for subnetworks to characterize the enhancement or suppression of the influence of input samples on each output neuron. Using WPA, we discover that neural network store and utilize information in a holographic way, that is, subnetworks encode all training samples in a coherent structure and thus only by investigating the weight pathways can one explore samples stored in the network. Furthermore, with WPA, we reveal fundamental learning modes of a neural network: the linear learning mode and the nonlinear learning mode. The former extracts linearly separable features while the latter extracts linearly inseparable features. The hidden-layer neurons self-organize into different classes for establishing learning modes and for reaching the training goal. The finding of learning modes provides us the theoretical ground for understanding some of the fundamental problems of machine learning, such as the dynamics of learning process, the role of linear and nonlinear neurons, as well as the role of network width and depth.

cs.LG↗

arXiv2022

PackCache: An Online Cost-driven Data Caching Algorithm in the Cloud

In this paper, we study a data caching problem in the cloud environment, where multiple frequently co-utilised data items could be packed as a single item being transferred to serve a sequence of data requests dynamically with reduced cost. To this end, we propose an online algorithm with respect to a homogeneous cost model, called PackCache, that can leverage the FP-Tree technique to mine those frequently co-utilised data items for packing whereby the incoming requests could be cost-effectively served online by exploiting the concept of anticipatory caching. We show the algorithm is 2αcompetitive, reaching the lower bound of the competitive ratio for any deterministic online algorithm on the studied caching problem, and also time and space efficient to serve the requests. Finally, we evaluate the performance of the algorithm via experimental studies to show its actual cost-effectiveness and scalability.

cs.DS↗

arXiv2022

Instability dynamics of nonlinear normal modes in the Fermi-Pasta-Ulam-Tsingou chains

Nonlinear normal modes are periodic orbits that survive in nonlinear chains, whose instability plays a crucial role in the dynamics of many-body Hamiltonian systems toward thermalization. Here we focus on how the stability of nonlinear modes depends on the perturbation strength and the system size to observe whether they have the same behavior in different models. To this end, as illustrating examples, the instability dynamics of the ${N}/{2}$ mode in both the Fermi-Pasta-Ulam-Tsingou (FPUT) -$α$ and -$β$ chains under fixed boundary conditions are studied systematically. Applying the Floquet theory, we show that for both models the stability time $T$ as a function of the perturbation strength $λ$ follows the same behavior; i.e., $T\propto(λ-λ_c)^{-\frac{1}{2}}$, where $λ_c$ is the instability threshold. The dependence of $λ_c$ on $N$ is also obtained. The results of $T$ and $λ_c$ agree well with those obtained by the direct molecular dynamics simulations. Finally, the effect of instability dynamics on the thermalization properties of a system is briefly discussed.

nlin.CD↗

arXiv2022

Thermalization of two- and three-dimensional classical lattices

Whether and how a system reaches thermalization is a fundamental issue of statistical physics. While for one-dimensional lattices this issue has been intensively studied in terms of energy equipartition for more than half a century, few work has been performed in the case of two- and three-dimensional lattices, and thus the thermalization dynamics remains unclear for more realistic lattices. In this Letter we investigate analytically and numerically the time-scaling of energy relaxation in these lattices. We show that the equipartition of energy is generally reached following a universal scheme for large enough lattices, regardless of its dimensionality, its specific lattice structure, and whether the system is translation invariant or not. Our results have practical significance in exploring the effect of high-order nonlinearities, i.e., the combining effect of multi-phonon process, in solid materials.

cond-mat.stat-mech↗