arXiv subjects

Hao Wu

Publications and source records attributed to Hao Wu.

At least 775 records · Page 43Linked to original sources

arXiv2019

Generation and Measurement of High-order Optical Vortex via Cross-Phase

The generation and measurement of optical vortex (OV) are the basis for a variety of related applications. However, the special case of high-order OVs has not been sufficient addressed yet. Herein, a generation and measurement method of high-order OV via utilizing the CP is investigated. In the experiment, we generate OVs with l=60, p=20 and successfully measure OVs with l=200,p=0, where experimental results agree well with simulation outcome. On this basis, the intensity distributions of LG and HG beams (corresponding to the generation and measurement) versus waist radius of initial light beams is discussed. This work provides an alternative method to generate or measure high-order OV, which will facilitate applications in optical micro-manipulation, quantum entanglement and rotation speed detection.

physics.optics↗

arXiv2019

Lithospheric delamination beneath the southern Puna plateau resolved by local earthquake tomography

We present a local earthquake tomography to illuminate the crustal and uppermost mantle structure beneath the southern Puna plateau and to test the delamination hypothesis. Vp and Vp/Vs ratios were obtained using travel time variations recorded by 75 temporary seismic stations between 2007 and 2009. In the upper crust, prominent low Vp anomalies are found beneath the main volcanic centers, indicating the presence of magma and melt beneath the southern Puna plateau. In the lowlands to the southeast of the Puna plateau, below the Sierras Pampeanas, a high Vp body is observed in the crust. Beneath the Moho at around 90 km depth, a strong high Vp anomaly is detected just west of the giant backarc Cerro Galan Ignimbrite caldera with the robustness of this feature being confirmed by multiple synthetic tests. This high velocity body can be interpreted as a delaminated block of lower crust and uppermost mantle lithosphere under the southern Puna plateau. The low velocities in the crust can be interpreted as having been induced by the delamination event that triggered the rise of fluids and melts into the crust and induced the high topography in this part of the plateau. The tomography also reveals low velocity anomalies that link arc magnetism at the Ojos del Salado volcanic center with slab seismicity clusters at depths of about 100 and 150 km and support fluid fluxing in the mantle wedge due to dehydration reaction within the subducted slab.

physics.geo-ph↗

arXiv2019

Structured Multi-Hashing for Model Compression

Despite the success of deep neural networks (DNNs), state-of-the-art models are too large to deploy on low-resource devices or common server configurations in which multiple models are held in memory. Model compression methods address this limitation by reducing the memory footprint, latency, or energy consumption of a model with minimal impact on accuracy. We focus on the task of reducing the number of learnable variables in the model. In this work we combine ideas from weight hashing and dimensionality reductions resulting in a simple and powerful structured multi-hashing method based on matrix products that allows direct control of model size of any deep network and is trained end-to-end. We demonstrate the strength of our approach by compressing models from the ResNet, EfficientNet, and MobileNet architecture families. Our method allows us to drastically decrease the number of variables while maintaining high accuracy. For instance, by applying our approach to EfficentNet-B4 (16M parameters) we reduce it to to the size of B0 (5M parameters), while gaining over 3% in accuracy over B0 baseline. On the commonly used benchmark CIFAR10 we reduce the ResNet32 model by 75% with no loss in quality, and are able to do a 10x compression while still achieving above 90% accuracy.

cs.LG↗

arXiv2019

Throughput Analysis in Cache-enabled Millimeter Wave HetNets with Access and Backhaul Integration

Recently, a mmWave-based access and backhaul integration heterogeneous networks (HetNets) architecture (mABHetNets) has been envisioned to provide high wireless capacity.Since the access link and the backhaul link share the same mmwave spectral resource, a large spectrum bandwidth is occupied by the backhaul link, which hinders the wireless access capacity improvement. To overcome the backhaul spectrum occupation problem and improve the network throughput in the existing mABHetNets, we introduce the cache at base stations (BSs). In detail, by caching popular files at small base stations (SBSs),mABHetNets can offload the backhaul link traffic and transfer the redundant backhaul spectrum to the access link to increase the network throughout. However, introducing cache in SBSs will also incur additional power consumption and reduce the transmission power, which can lower the network throughput.In this paper, we investigate spectrum partition between the access link and the backhaul link as well as cache allocation to improve the network throughput in mABHetNets. With the stochastic geometry tool, we develop an analytical framework to characterize cache-enabled mABHetNets and obtain the signal-to-interference-plus-noise ratio (SINR) distributions in line-of-sight (LoS) and non-line-of-sight (NLoS) paths. Then we utilize the SINR distribution to derive the average potential throughput(APT). Extensive numerical results show that introducing cache can bring up to 80% APT gain to the existing mABHetNets.

cs.NI↗

arXiv2019

Modeling and Performance Analysis in Cache-enabled Millimeter Wave HetNets with Access and Backhaul Integration

Recently, a mmWave-based access and backhaul integration HetNetsarchitecture (mABHetNets) has been envisioned to provide the high link capacity in both access and backhaul links. In mABHetNets, the access link and the backhaul link will share the same mm-wave spectral resources to serve cellular users as well as to maintain the backhaul capacity of the SBSs. However, a large spectrum bandwidth are occupied by the backhaul link to maintain the backhaul link capacity. Up to 50\% mmWave spectrum resource will be used in backhaul link to satisfy the high speed data traffic. Such "\emph{spectrum occupation}" of backhaul link has restricted network performance to achieve a better possible improvement. To overcome spectrum occupation problem in the existing mABHetNets, involving caches in the mABHetNets brings a opportunity. When the backhaul traffic is offloaded by caching popular fils in the SBSs and the corresponding mmWave spectrum can be transfered to the access link by the bandwidth partition. Then with more spectrum for access link, both the network throughout and the coverage ability are increased significantly. However, the caching power consumption will be increased and the EE will be decreased. Therefore, we analyze the througput and EE in mABHetNets by stochastic geometry tool. Extensive numerical and simulations results are carried out to validate our work.

cs.NI↗

arXiv2019

Separation property and convergence to equilibrium for the equation and dynamic boundary condition of Cahn-Hilliard type with singular potential

We consider a class of Cahn-Hilliard equation that models phase separation process of binary mixtures involving nontrivial boundary interactions in a bounded domain with non-permeable wall. The system is characterized by certain dynamic type boundary conditions and the total mass, in the bulk and on the boundary, is conserved for all time. For the case with physically relevant singular (e.g., logarithmic) potential, global regularity of weak solutions is established. In particular, when the spatial dimension is two, we show the instantaneous strict separation property such that for arbitrary positive time any weak solution stays away from the pure phases +1 and -1, while in the three dimensional case, an eventual separation property for large time is obtained. As a consequence, we prove that every global weak solution converges to a single equilibrium as the time goes to infinity, by the usage of an extended Lojasiewicz-Simon inequality.

math.AP↗

arXiv2019

Sub-Channel Allocation for Device-to-Device Underlaying Full-Duplex mmWave Small Cells using Coalition Formation Games

The small cells in millimeter wave (mmWave) have been utilized extensively for the fifth generation (5G) mobile networks. And full-duplex (FD) communications become possible with the development of self-interference (SI) cancellation technology. In this paper, we focus on the optimal allocation of the sub-channels in the small cells deployed densely underlying the mmWave networks. In order to improve the resource utilization and network capacity, device-to-device (D2D) communications are adopted in networks. For maximizing the system throughput, we propose a cooperative sub-channel allocation approach based on coalition formation game, and theoretically prove the convergence of the proposed allocation algorithm. The utility of the coalition formation game is closely related to the sum system throughput in this paper, and each link makes an individual decision to leave or join any coalition based on the principle of maximizing the utility. Besides, a threshold of the minimum transmission rate is given to improve the system fairness. Through extensive simulations, we demonstrate that our proposed scheme has a near-optimal performance on sum throughput. In addition, we verify the low complexity of the proposed scheme by simulating the number of switch operations.

cs.NI↗

arXiv2019

On the convergence of FK-Ising Percolation to SLE$(16/3, 16/3-6)$

We give a simplified and complete proof of the convergence of the chordal exploration process in critical FK-Ising percolation to chordal SLE$_κ( κ-6)$ with $κ=16/3$. Our proof follows the classical excursion-construction of SLE$_κ(κ-6)$ processes in the continuum and we are thus led to introduce suitable cut-off stopping times in order to analyse the behaviour of the driving function of the discrete system when Dobrushin boundary condition collapses to a single point. Our proof is very different from [KS15, KS16] as it only relies on the convergence to the chordal SLE$_κ$ process in Dobrushin boundary condition and does not require the introduction of a new observable. Still, it relies crucially on several ingredients: a) the powerful topological framework developed in [KS17] as well as its follow-up paper [CDCH$^+$14], b) the strong RSW Theorem from [CDCH16], c) the proof is inspired from the appendix A in [BH16]. One important emphasis of this paper is to carefully write down some properties which are often considered {\em folklore} in the literature but which are only justified so far by hand-waving arguments. The main examples of these are: 1) the convergence of natural discrete stopping times to their continuous analogues. (The usual hand-waving argument destroys the spatial Markov property). 2) the fact that the discrete spatial Markov property is preserved in the the scaling limit. (The enemy being that $\mathbb{E}[X_n |\, Y_n]$ does not necessarily converge to $\mathbb{E}[X|\, Y]$ when $(X_n,Y_n)\to (X,Y)$). We end the paper with a detailed sketch of the convergence to radial SLE$_κ( κ-6)$ when $κ=16/3$ as well as the derivation of Onsager's one-arm exponent $1/8$.

math.PR↗

arXiv2019

Electrically controlled localized charge trapping at amorphous fluoropolymer-electrolyte interfaces

Charge trapping is a long-standing problem in electrowetting-on-dielectric (EWOD), causing reliability reduction and restricting its practical applications. Although this phenomenon has been investigated macroscopically, the microscopic investigations are still lacking. In this work, the trapped charges are proven to be localized at three-phase contact line region by using three detecting methods -- local contact angle measurements, electrowetting (EW) probe, and Kelvin Probe Force Microscopy (KPFM). Moreover, we demonstrate that this EW-induced charge trapping phenomenon can be utilized as a simple and low-cost method to deposit charges on fluoropolymer surfaces. Charge density near the three-phase contact line up to 0.46 mC/m2 and the line width with deposited charges ranging from 20 to 300 micrometer are achieved by the proposed method. Particularly, negative charge densities do not degrade even after harsh testing with a water droplet on top of the sample surfaces for 12 hours, as well as after being treated by water vapor for 3 hours. These findings provide an approach for applications which desire stable and controllable surface charges.

physics.app-ph↗

arXiv2019

On a Class of Sixth-order Cahn-Hilliard Type Equations with Logarithmic Potential

We consider a class of six-order Cahn-Hilliard equations with logarithmic type potential. This system is closely connected with some important phase-field models relevant in different applications, for instance, the functionalized Cahn-Hilliard equation that describes phase separation in mixtures of amphiphilic molecules in solvent, and the Willmore regularization of Cahn-Hilliard equation for anisotropic crystal and epitaxial growth. The singularity of the configuration potential guarantees that the solution always stays in the physical relevant domain [-1,1]. Meanwhile, the resulting system is characterized by some highly singular diffusion terms that make the mathematical analysis more involved. We prove existence and uniqueness of global weak solutions and show their parabolic regularization property for any positive time. Besides, we investigate long-time behavior of the system, proving existence of the global attractor for the associated dynamical process in a suitable complete metric space.

math.AP↗

arXiv2019

Online Optimization of Wireless Powered Mobile-Edge Computing for Heterogeneous Industrial Internet of Things

A spurt of progress in wireless power transfer (WPT) and mobile edge computing (MEC) provides a promising approach for Industrial Internet of Things (IIoT) to enhance the quality and productivity of manufacturing. Scheduling in such a scenario is challenging due to congested wireless channels, time-dependent energy constraints, complicated device heterogeneity, and prohibitive signaling overheads. In this paper, we first propose an online algorithm, called energy-aware resource scheduling (ERS), to maximize the system utility comprising throughput and fairness, with consideration on both system sustainability and stability. Based on Lyapunov optimization and convex optimization techniques, the proposed algorithm achieves asymptotic optimality for heterogeneous IIoT systems without prior knowledge of network state information (NSI). Subsequently, we extend the ERS algorithm to a more realistic scenario where the overhead and delay of NSI feedbacks are nonnegligible. The optimal scheduling decisions of the scenario are provided, and the optimality loss on system utility under outdated NSI is analyzed. Simulations verify our theoretical claims and demonstrate the gains of our proposed ERS algorithm over alternative benchmark schemes.

eess.SP↗

arXiv2019

Learning Latent Parameters without Human Response Patterns: Item Response Theory with Artificial Crowds

Incorporating Item Response Theory (IRT) into NLP tasks can provide valuable information about model performance and behavior. Traditionally, IRT models are learned using human response pattern (RP) data, presenting a significant bottleneck for large data sets like those required for training deep neural networks (DNNs). In this work we propose learning IRT models using RPs generated from artificial crowds of DNN models. We demonstrate the effectiveness of learning IRT models using DNN-generated data through quantitative and qualitative analyses for two NLP tasks. Parameters learned from human and machine RPs for natural language inference and sentiment analysis exhibit medium to large positive correlations. We demonstrate a use-case for latent difficulty item parameters, namely training set filtering, and show that using difficulty to sample training data outperforms baseline methods. Finally, we highlight cases where human expectation about item difficulty does not match difficulty as estimated from the machine RPs.

cs.CL↗

arXiv2019

An encoding framework with brain inner state for natural image identification

Neural encoding and decoding, which aim to characterize the relationship between stimuli and brain activities, have emerged as an important area in cognitive neuroscience. Traditional encoding models, which focus on feature extraction and mapping, consider the brain as an input-output mapper without inner states. In this work, inspired by the fact that human brain acts like a state machine, we proposed a novel encoding framework that combines information from both the external world and the inner state to predict brain activity. The framework comprises two parts: forward encoding model that deals with visual stimuli and inner state model that captures influence from intrinsic connections in the brain. The forward model can be any traditional encoding model, making the framework flexible. The inner state model is a linear model to utilize information in the prediction residuals of the forward model. The proposed encoding framework can achieve much better performance on natural image identification from fMRI response than forwardonly models. The identification accuracy will decrease slightly with the dataset size increasing, but remain relatively stable with different identification methods. The results confirm that the new encoding framework is effective and robust when used for brain decoding.

q-bio.NC↗

arXiv2019

Variational approach for learning Markov processes from time series data

Inference, prediction and control of complex dynamical systems from time series is important in many areas, including financial markets, power grid management, climate and weather modeling, or molecular dynamics. The analysis of such highly nonlinear dynamical systems is facilitated by the fact that we can often find a (generally nonlinear) transformation of the system coordinates to features in which the dynamics can be excellently approximated by a linear Markovian model. Moreover, the large number of system variables often change collectively on large time- and length-scales, facilitating a low-dimensional analysis in feature space. In this paper, we introduce a variational approach for Markov processes (VAMP) that allows us to find optimal feature mappings and optimal Markovian models of the dynamics from given time series data. The key insight is that the best linear model can be obtained from the top singular components of the Koopman operator. This leads to the definition of a family of score functions called VAMP-r which can be calculated from data, and can be employed to optimize a Markovian model. In addition, based on the relationship between the variational scores and approximation errors of Koopman operators, we propose a new VAMP-E score, which can be applied to cross-validation for hyper-parameter optimization and model selection in VAMP. VAMP is valid for both reversible and nonreversible processes and for stationary and non-stationary processes or realizations.

stat.ML↗

arXiv2019

Detecting 11K Classes: Large Scale Object Detection without Fine-Grained Bounding Boxes

Recent advances in deep learning greatly boost the performance of object detection. State-of-the-art methods such as Faster-RCNN, FPN and R-FCN have achieved high accuracy in challenging benchmark datasets. However, these methods require fully annotated object bounding boxes for training, which are incredibly hard to scale up due to the high annotation cost. Weakly-supervised methods, on the other hand, only require image-level labels for training, but the performance is far below their fully-supervised counterparts. In this paper, we propose a semi-supervised large scale fine-grained detection method, which only needs bounding box annotations of a smaller number of coarse-grained classes and image-level labels of large scale fine-grained classes, and can detect all classes at nearly fully-supervised accuracy. We achieve this by utilizing the correlations between coarse-grained and fine-grained classes with shared backbone, soft-attention based proposal re-ranking, and a dual-level memory module. Experiment results show that our methods can achieve close accuracy on object detection to state-of-the-art fully-supervised methods on two large scale datasets, ImageNet and OpenImages, with only a small fraction of fully annotated classes.

cs.CV↗

arXiv2019

On a transmission problem for equation and dynamic boundary condition of Cahn-Hilliard type with nonsmooth potentials

This paper is concerned with well-posedness of the Cahn-Hilliard equation subject to a class of new dynamic boundary conditions. The system was recently derived in Liu-Wu (Arch. Ration. Mech. Anal. 233 (2019), 167-247) via an energetic variational approach and it naturally fulfills three physical constraints such as mass conservation, energy dissipation and force balance. The target problem examined in this paper can be viewed as a transmission problem that consists of Cahn-Hilliard type equations both in the bulk and on the boundary. In our approach, we are able to deal with a general class of potentials with double-well structure, including the physically relevant logarithmic potential and the non-smooth double-obstacle potential. Existence, uniqueness and continuous dependence of global weak solutions are established. The proof is based on a novel time-discretization scheme for the approximation of the continuous problem. Besides, a regularity result is shown with the aim of obtaining a strong solution to the system.

math.AP↗

arXiv2019

Boltzmann Generators -- Sampling Equilibrium States of Many-Body Systems with Deep Learning

Computing equilibrium states in condensed-matter many-body systems, such as solvated proteins, is a long-standing challenge. Lacking methods for generating statistically independent equilibrium samples in "one shot", vast computational effort is invested for simulating these system in small steps, e.g., using Molecular Dynamics. Combining deep learning and statistical mechanics, we here develop Boltzmann Generators, that are shown to generate unbiased one-shot equilibrium samples of representative condensed matter systems and proteins. Boltzmann Generators use neural networks to learn a coordinate transformation of the complex configurational equilibrium distribution to a distribution that can be easily sampled. Accurate computation of free energy differences and discovery of new configurations are demonstrated, providing a statistical mechanics tool that can avoid rare events during sampling without prior knowledge of reaction coordinates.

stat.ML↗

arXiv2019

Wasserstein-Fisher-Rao Document Distance

As a fundamental problem of natural language processing, it is important to measure the distance between different documents. Among the existing methods, the Word Mover's Distance (WMD) has shown remarkable success in document semantic matching for its clear physical insight as a parameter-free model. However, WMD is essentially based on the classical Wasserstein metric, thus it often fails to robustly represent the semantic similarity between texts of different lengths. In this paper, we apply the newly developed Wasserstein-Fisher-Rao (WFR) metric from unbalanced optimal transport theory to measure the distance between different documents. The proposed WFR document distance maintains the great interpretability and simplicity as WMD. We demonstrate that the WFR document distance has significant advantages when comparing the texts of different lengths. In addition, an accelerated Sinkhorn based algorithm with GPU implementation has been developed for the fast computation of WFR distances. The KNN classification results on eight datasets have shown its clear improvement over WMD.

cs.LG↗