arXiv subjects

Hang Zhang

Publications and source records attributed to Hang Zhang.

At least 235 records · Page 13Linked to original sources

arXiv2019

Machine Learning for Novel Thermal-Materials Discovery: Early Successes, Opportunities, and Challenges

High-throughput computational and experimental design of materials aided by machine learning have become an increasingly important field in material science. This area of research has emerged in leaps and bounds in the thermal sciences, in part due to the advances in computational and experimental methods in obtaining thermal properties of materials. In this paper, we provide a current overview of some of the recent work and highlight the challenges and opportunities that are ahead of us in this field. In particular, we focus on the use of machine learning and high-throughput methods for screening of thermal conductivity for compounds, composites and alloys as well as interfacial thermal conductance. These new tools have brought about a feedback mechanism for understanding new correlations and identifying new descriptors, speeding up the discovery of novel thermal functional materials.

cond-mat.mtrl-sci↗

arXiv2018

Bag of Tricks for Image Classification with Convolutional Neural Networks

Much of the recent progress made in image classification research can be credited to training procedure refinements, such as changes in data augmentations and optimization methods. In the literature, however, most refinements are either briefly mentioned as implementation details or only visible in source code. In this paper, we will examine a collection of such refinements and empirically evaluate their impact on the final model accuracy through ablation study. We will show that, by combining these refinements together, we are able to improve various CNN models significantly. For example, we raise ResNet-50's top-1 validation accuracy from 75.3% to 79.29% on ImageNet. We will also demonstrate that improvement on image classification accuracy leads to better transfer learning performance in other application domains such as object detection and semantic segmentation.

cs.CV↗

arXiv2018

The swimming of a deforming helix

Many microorganisms and artificial microswimmers use helical appendages in order to generate locomotion. Though often rotated so as to produce thrust, some species of bacteria such Spiroplasma, Rhodobacter sphaeroides and Spirochetes induce movement by deforming a helical-shaped body. Recently, artificial devices have been created which also generate motion by deforming their helical body in a non-reciprocal way (Mourran et al., Adv. Mater., 29, 1604825, 2017). Inspired by these systems, we investigate the transport of a deforming helix within a viscous fluid. Specifically, we consider a swimmer that maintains a helical centreline and a single handedness while changing its helix radius, pitch and wavelength uniformly across the body. We first discuss how a deforming helix can create a non-reciprocal translational and rotational swimming stroke and identify its principle direction of motion. We then determine the leading-order physics for helices with small helix radius before considering the general behaviour for different configuration parameters and how these swimmers can be optimised. Finally, we explore how the presence of walls, gravity, and defects in the centreline allow the helical device to break symmetries, increase its speed, and generate transport in directions not available to helices in bulk fluids.

physics.bio-ph↗

arXiv2018

Temperature effect on the coupling between coherent longitudinal phonons and plasmons in n- and p-type GaAs

The coupling between longitudinal optical (LO) phonons and plasmons plays a fundamental role in determining the performance of doped semiconductor devices. In this work, we report a comparative investigation into the dependence of the coupling on temperature and doping in n- and p-type GaAs by using ultrafast optical phonon spectroscopy. A suppression of coherent oscillations has been observed in p-type GaAs at lower temperature, strikingly different from n-type GaAs and other materials in which coherent oscillations are strongly enhanced by cooling. We attribute this unexpected observation to a cooling-induced elongation of the depth of the depletion layer which effectively increases the screening time of surface field due to a slow diffusion of photoexcited carriers in p-type GaAs. Such an increase breaks the requirement for the generation of coherent LO phonons and, in turn, LO phonon-plasmon coupled modes because of their delayed formation in time.

cond-mat.mtrl-sci↗

arXiv2018

Deep Texture Manifold for Ground Terrain Recognition

We present a texture network called Deep Encoding Pooling Network (DEP) for the task of ground terrain recognition. Recognition of ground terrain is an important task in establishing robot or vehicular control parameters, as well as for localization within an outdoor environment. The architecture of DEP integrates orderless texture details and local spatial information and the performance of DEP surpasses state-of-the-art methods for this task. The GTOS database (comprised of over 30,000 images of 40 classes of ground terrain in outdoor scenes) enables supervised recognition. For evaluation under realistic conditions, we use test images that are not from the existing GTOS dataset, but are instead from hand-held mobile phone videos of similar terrain. This new evaluation dataset, GTOS-mobile, consists of 81 videos of 31 classes of ground terrain such as grass, gravel, asphalt and sand. The resultant network shows excellent performance not only for GTOS-mobile, but also for more general databases (MINC and DTD). Leveraging the discriminant features learned from this network, we build a new texture manifold called DEP-manifold. We learn a parametric distribution in feature space in a fully supervised manner, which gives the distance relationship among classes and provides a means to implicitly represent ambiguous class boundaries. The source code and database are publicly available.

cs.CV↗

arXiv2018

Context Encoding for Semantic Segmentation

Recent work has made significant progress in improving spatial resolution for pixelwise labeling with Fully Convolutional Network (FCN) framework by employing Dilated/Atrous convolution, utilizing multi-scale features and refining boundaries. In this paper, we explore the impact of global contextual information in semantic segmentation by introducing the Context Encoding Module, which captures the semantic context of scenes and selectively highlights class-dependent featuremaps. The proposed Context Encoding Module significantly improves semantic segmentation results with only marginal extra computation cost over FCN. Our approach has achieved new state-of-the-art results 51.7% mIoU on PASCAL-Context, 85.9% mIoU on PASCAL VOC 2012. Our single model achieves a final score of 0.5567 on ADE20K test set, which surpass the winning entry of COCO-Place Challenge in 2017. In addition, we also explore how the Context Encoding Module can improve the feature representation of relatively shallow networks for the image classification on CIFAR-10 dataset. Our 14 layer network has achieved an error rate of 3.45%, which is comparable with state-of-the-art approaches with over 10 times more layers. The source code for the complete system are publicly available.

cs.CV↗

arXiv2017

Multi-style Generative Network for Real-time Transfer

Despite the rapid progress in style transfer, existing approaches using feed-forward generative network for multi-style or arbitrary-style transfer are usually compromised of image quality and model flexibility. We find it is fundamentally difficult to achieve comprehensive style modeling using 1-dimensional style embedding. Motivated by this, we introduce CoMatch Layer that learns to match the second order feature statistics with the target styles. With the CoMatch Layer, we build a Multi-style Generative Network (MSG-Net), which achieves real-time performance. We also employ an specific strategy of upsampled convolution which avoids checkerboard artifacts caused by fractionally-strided convolution. Our method has achieved superior image quality comparing to state-of-the-art approaches. The proposed MSG-Net as a general approach for real-time style transfer is compatible with most existing techniques including content-style interpolation, color-preserving, spatial control and brush stroke size control. MSG-Net is the first to achieve real-time brush-size control in a purely feed-forward manner for style transfer. Our implementations and pre-trained models for Torch, PyTorch and MXNet frameworks will be publicly available.

cs.CV↗

arXiv2017

Differential Angular Imaging for Material Recognition

Material recognition for real-world outdoor surfaces has become increasingly important for computer vision to support its operation "in the wild." Computational surface modeling that underlies material recognition has transitioned from reflectance modeling using in-lab controlled radiometric measurements to image-based representations based on internet-mined images of materials captured in the scene. We propose to take a middle-ground approach for material recognition that takes advantage of both rich radiometric cues and flexible image capture. We realize this by developing a framework for differential angular imaging, where small angular variations in image capture provide an enhanced appearance representation and significant recognition improvement. We build a large-scale material database, Ground Terrain in Outdoor Scenes (GTOS) database, geared towards real use for autonomous agents. The database consists of over 30,000 images covering 40 classes of outdoor ground terrain under varying weather and lighting conditions. We develop a novel approach for material recognition called a Differential Angular Imaging Network (DAIN) to fully leverage this large dataset. With this novel network architecture, we extract characteristics of materials encoded in the angular and spatial gradients of their appearance. Our results show that DAIN achieves recognition performance that surpasses single view or coarsely quantized multiview images. These results demonstrate the effectiveness of differential angular imaging as a means for flexible, in-place material recognition.

cs.CV↗

arXiv2017

Photo-realistic Facial Texture Transfer

Style transfer methods have achieved significant success in recent years with the use of convolutional neural networks. However, many of these methods concentrate on artistic style transfer with few constraints on the output image appearance. We address the challenging problem of transferring face texture from a style face image to a content face image in a photorealistic manner without changing the identity of the original content image. Our framework for face texture transfer (FaceTex) augments the prior work of MRF-CNN with a novel facial semantic regularization that incorporates a face prior regularization smoothly suppressing the changes around facial meso-structures (e.g eyes, nose and mouth) and a facial structure loss function which implicitly preserves the facial structure so that face texture can be transferred without changing the original identity. We demonstrate results on face images and compare our approach with recent state-of-the-art methods. Our results demonstrate superior texture transfer because of the ability to maintain the identity of the original face image.

cs.CV↗

arXiv2016

Deep TEN: Texture Encoding Network

We propose a Deep Texture Encoding Network (Deep-TEN) with a novel Encoding Layer integrated on top of convolutional layers, which ports the entire dictionary learning and encoding pipeline into a single model. Current methods build from distinct components, using standard encoders with separate off-the-shelf features such as SIFT descriptors or pre-trained CNN features for material recognition. Our new approach provides an end-to-end learning framework, where the inherent visual vocabularies are learned directly from the loss function. The features, dictionaries and the encoding representation for the classifier are all learned simultaneously. The representation is orderless and therefore is particularly useful for material and texture recognition. The Encoding Layer generalizes robust residual encoders such as VLAD and Fisher Vectors, and has the property of discarding domain specific information which makes the learned convolutional features easier to transfer. Additionally, joint training using multiple datasets of varied sizes and class labels is supported resulting in increased recognition performance. The experimental results show superior performance as compared to state-of-the-art methods using gold-standard databases such as MINC-2500, Flickr Material Database, KTH-TIPS-2b, and two recent databases 4D-Light-Field-Material and GTOS. The source code for the complete system are publicly available.

cs.CV↗

arXiv2016

Future Wireless Network: MyNET Platform and End-to-End Network Slicing

Future wireless networks are facing new challenges. These new challenges require new solutions and strategies of the network deployment, management, and operation. Many driving factors are decisive in the re-definition and re-design of the future wireless network architecture. In the previously published paper "5G Wireless Network - MyNET and SONAC", MyNET and SONAC, a future network architecture, are described. This paper elaborates MyNET platform with more details. The design principles of MyNET architecture, the development of MyNET platform, the functions of MyNET platform, the automation of the creation and the management of end-to-end slices by MyNET, and the new capabilities enabled by MyNET are described in details.

cs.NI↗

arXiv2016

Robust Matrix Regression

Modern technologies are producing datasets with complex intrinsic structures, and they can be naturally represented as matrices instead of vectors. To preserve the latent data structures during processing, modern regression approaches incorporate the low-rank property to the model and achieve satisfactory performance for certain applications. These approaches all assume that both predictors and labels for each pair of data within the training set are accurate. However, in real-world applications, it is common to see the training data contaminated by noises, which can affect the robustness of these matrix regression methods. In this paper, we address this issue by introducing a novel robust matrix regression method. We also derive efficient proximal algorithms for model training. To evaluate the performance of our methods, we apply it to real world applications with comparative studies. Our method achieves the state-of-the-art performance, which shows the effectiveness and the practical value of our method.

cs.LG↗

arXiv2016

mRNA-miRNA Reciprocal Regulation Enabled Bistable Switch Directs Cell Fate Decision

miRNAs serve as crucial post-transcriptional regulators in various essential cell fate decision. However, the contribution of the mRNA-miRNA mutual regulation to bistability is not fully understood. Here, we built a set of mathematical models of mRNA-miRNA interactions and systematically analyzed the sensitivity of response curves under various conditions. First, we found that mRNA-miRNA reciprocal regulation could manifest ultrasensitivity to subserve the generation of bistability when equipped with a positive feedback loop. Second, the region of bistability is expanded by a stronger competing mRNA (ceRNA). Interesting, bistability can be emerged without feedback loop if multiple miRNA binding sites exist on a target mRNA. Thus, we demonstrated the importance of simple mRNA-miRNA reciprocal regulation in cell fate decision.

q-bio.MN↗

arXiv2016

Future Network: End-to-End Slicing and Hop-On (a Slice)

The concept of network slice, i.e.,a service customized virtual network (VN) is attracting more and more attentions in the telecommunication industry. A slice is a set of network resources which fits the service attributes and requirements of customer services. The network resources consist of cloud resources and communication link resources. A slice can serve one or more customer services which share the similar service attributes and requirements. To define, create and manage a slice (VN) is one aspect of future networks. Another aspect is the slice operation, i.e., the provisioning of services to customers using created slices. In this paper, the focus is put on the configuration of slices and the operation of slices. In this paper, the detailed description of slice (VN) configuration is provided. A new concept of hop-on (a slice) is described. Given a well defined and configured end-to-end slice (VN), the realtime data traffic delivery over a slice is governed by network operation control entities, which are also pre-configured. Therefore, the procedure of customer traffic delivery over a slice is just like a traveler hopping on tourist bus and then the traffic control officers at key intersections directing the traveler to go through pre-designed routes until the destination is reached.

cs.NI↗

arXiv2016

Friction from Reflectance: Deep Reflectance Codes for Predicting Physical Surface Properties from One-Shot In-Field Reflectance

Images are the standard input for vision algorithms, but one-shot infield reflectance measurements are creating new opportunities for recognition and scene understanding. In this work, we address the question of what reflectance can reveal about materials in an efficient manner. We go beyond the question of recognition and labeling and ask the question: What intrinsic physical properties of the surface can be estimated using reflectance? We introduce a framework that enables prediction of actual friction values for surfaces using one-shot reflectance measurements. This work is a first of its kind vision-based friction estimation. We develop a novel representation for reflectance disks that capture partial BRDF measurements instantaneously. Our method of deep reflectance codes combines CNN features and fisher vector pooling with optimal binary embedding to create codes that have sufficient discriminatory power and have important properties of illumination and spatial invariance. The experimental results demonstrate that reflectance can play a new role in deciphering the underlying physical properties of real-world scenes.

cs.CV↗

arXiv2016

Achieving Diverse and Monoallelic Olfactory Receptor Selection Through Dual-Objective Optimization Design

Multiple-objective optimization is common in biological systems. In the mammalian olfactory system, each sensory neuron stochastically expresses only one out of up to thousands of olfactory receptor (OR) gene alleles; at organism level the types of expressed ORs need to be maximized. Existing models focus only on monoallele activation, and cannot explain recent observations in mutants, especially the reduced global diversity of expressed ORs in G9a/GLP knockouts. In this work we integrated existing information on OR expression, and constructed a comprehensive model that has all its components based on physical interactions. Analyzing the model reveals an evolutionarily optimized three-layer regulation mechanism, which includes zonal segregation, epigenetic barrier crossing coupled to a negative feedback loop that mechanistically differs from previous theoretical proposals, and a previously unidentified enhancer competition step. This model not only recapitulates monoallelic OR expression, but also elucidates how the olfactory system maximizes and maintains the diversity of OR expression, and has multiple predictions validated by existing experimental results. Through making analogy to a physical system with thermally activated barrier crossing and comparative reverse engineering analyses, the study reveals that the olfactory receptor selection system is optimally designed, and particularly underscores cooperativity and synergy as a general design principle for multi-objective optimization in biology.

q-bio.MN↗

arXiv2016

Frequency-Domain Response Based Timing Synchronization: A Near Optimal Sampling Phase Criterion for TDS-OFDM

In time-domain synchronous OFDM (TDS-OFDM) system for digital television terrestrial multimedia broadcasting (DTMB) standard, the baseband OFDM signal is upsampled and shaping filtered by square root raised cosine (SRRC) filter before digital-to-analog converter (DAC). Much of the work in the area of timing synchronization for TDS-OFDM focuses on frame synchronization and sampling clock frequency offset recovery, which does not consider the sampling clock phase offset due to the upsampling and SRRC filter. This paper evaluates the bit-error-rate (BER) effect of sampling clock phase offset in TDS-OFDM system. First, we provide the BER for M-order quadrature amplitude modulation (M-QAM) in uncoded TDS-OFDM system. Second, under the condition of the optimal BER criterion and additive white Gaussian noise (AWGN) channel, we propose a near optimal sampling phase estimation criterion based on frequency-domain response. Simulations demonstrate that the proposed criterion also has good performance in actual TDS-OFDM system with channel coding over multipath channels, and it is superior to the conventional symbol timing recovery methods for TDS-OFDM system.

cs.IT↗

arXiv2015

Temperature Dependent Mean Free Path Spectra of Thermal Phonons Along the c-axis of Graphite

Heat conduction in graphite has been studied for decades because of its exceptionally large thermal anisotropy. While the bulk thermal conductivities along the in-plane and cross-plane directions are well known, less understood are the microscopic properties of the thermal phonons responsible for heat conduction. In particular, recent experimental and computational works indicate that the average phonon mean free path (MFP) along the c-axis is considerably larger than that estimated by kinetic theory, but the distribution of MFPs remains unknown. Here, we report the first quantitative measurements of c-axis phonon MFP spectra in graphite at a variety of temperatures using time-domain thermoreflectance measurements of graphite flakes with variable thickness. Our results indicate that c-axis phonon MFPs have values of a few hundred nanometers at room temperature and a much narrower distribution than in isotropic crystals. At low temperatures, phonon scattering is dominated by grain boundaries separating crystalline regions of different rotational orientation. Our study provides important new insights into heat transport and phonon scattering mechanisms in graphite and other anisotropic van der Waals solids.

cond-mat.mes-hall↗