arXiv subjects

Chao Wang

Publications and source records attributed to Chao Wang.

At least 703 records · Page 39Linked to original sources

arXiv2017

CirCNN: Accelerating and Compressing Deep Neural Networks Using Block-CirculantWeight Matrices

Large-scale deep neural networks (DNNs) are both compute and memory intensive. As the size of DNNs continues to grow, it is critical to improve the energy efficiency and performance while maintaining accuracy. For DNNs, the model size is an important factor affecting performance, scalability and energy efficiency. Weight pruning achieves good compression ratios but suffers from three drawbacks: 1) the irregular network structure after pruning; 2) the increased training complexity; and 3) the lack of rigorous guarantee of compression ratio and inference accuracy. To overcome these limitations, this paper proposes CirCNN, a principled approach to represent weights and process neural networks using block-circulant matrices. CirCNN utilizes the Fast Fourier Transform (FFT)-based fast multiplication, simultaneously reducing the computational complexity (both in inference and training) from O(n2) to O(nlogn) and the storage complexity from O(n2) to O(n), with negligible accuracy loss. Compared to other approaches, CirCNN is distinct due to its mathematical rigor: it can converge to the same effectiveness as DNNs without compression. The CirCNN architecture, a universal DNN inference engine that can be implemented on various hardware/software platforms with configurable network architecture. To demonstrate the performance and energy efficiency, we test CirCNN in FPGA, ASIC and embedded processors. Our results show that CirCNN architecture achieves very high energy efficiency and performance with a small hardware footprint. Based on the FPGA implementation and ASIC synthesis results, CirCNN achieves 6-102X energy efficiency improvements compared with the best state-of-the-art results.

cs.CV↗

arXiv2017

High Efficient Reconstruction of Single-shot T2 Mapping from OverLapping-Echo Detachment Planar Imaging Based on Deep Residual Network

Purpose: An end-to-end deep convolutional neural network (CNN) based on deep residual network (ResNet) was proposed to efficiently reconstruct reliable T2 mapping from single-shot OverLapping-Echo Detachment (OLED) planar imaging. Methods: The training dataset was obtained from simulations carried out on SPROM software developed by our group. The relationship between the original OLED image containing two echo signals and the corresponded T2 mapping was learned by ResNet training. After the ResNet was trained, it was applied to reconstruct the T2 mapping from simulation and in vivo human brain data. Results: Though the ResNet was trained entirely on simulated data, the trained network was generalized well to real human brain data. The results from simulation and in vivo human brain experiments show that the proposed method significantly outperformed the echo-detachment-based method. Reliable T2 mapping was achieved within tens of milliseconds after the network had been trained while the echo-detachment-based OLED reconstruction method took minutes. Conclusion: The proposed method will greatly facilitate real-time dynamic and quantitative MR imaging via OLED sequence, and ResNet has the potential to reconstruct images from complex MRI sequence efficiently.

cs.CV↗

arXiv2017

Global Analysis of Charmless $B$ Decays into Two Vector Mesons in Soft-Collinear Effective Theory

Under the framework of soft-collinear effective theory, we analyze the charmless $B\to VV$ decays in a global way at leading power in $1/m_b$ and leading order in $α_s$ with $V$ denoting a light vector meson. In the flavor SU(3) symmetry, decay amplitudes for the 28 decay modes are expressed in terms of 8 nonperturbative parameters. With 35 experimental results, we fit these 8 nonperturbative parameters. Annihilation contributions are neglected due to power suppression in the $m_b\to \infty$ limit, so we include in the fit the nonperturbative charm penguins, which will play an important role in understanding the direct CP asymmetries. Charming penguins are also responsible for the large transverse polarizations of penguin-dominated and color-suppressed decays. With the best fitted parameters, we calculate all possible physical observables of 28 decay modes, including branching fractions, direct CP asymmetries, and the complete set of polarization observables. Most of our results are compatible with the present experimental data when available, while the others can be examined on the ongoing LHCb experiment and the forthcoming Belle-II experiment. Moreover, the agreements and differences with results in QCD factorization and perturbative QCD approach are also discussed. A few observables are suggested to discriminate these different approaches.

hep-ph↗

arXiv2017

TNSPackage: A Fortran2003 library designed for tensor network state methods

Recently, the tensor network states (TNS) methods have proven to be very powerful tools to investigate the strongly correlated many-particle physics in one and two dimensions. The implementation of TNS methods depends heavily on the operations of tensors, including contraction, permutation, reshaping tensors, SVD and son on. Unfortunately, the most popular computer languages for scientific computation, such as Fortran and C/C++ do not have a standard library for such operations, and therefore make the coding of TNS very tedious. We develop a Fortran2003 package that includes all kinds of basic tensor operations designed for TNS. It is user-friendly and flexible for different forms of TNS, and therefore greatly simplifies the coding work for the TNS methods.

cond-mat.str-el↗

arXiv2017

Bayesian Realized-GARCH Models for Financial Tail Risk Forecasting Incorporating Two-sided Weibull Distribution

The realized GARCH framework is extended to incorporate the two-sided Weibull distribution, for the purpose of volatility and tail risk forecasting in a financial time series. Further, the realized range, as a competitor for realized variance or daily returns, is employed in the realized GARCH framework. Further, sub-sampling and scaling methods are applied to both the realized range and realized variance, to help deal with inherent micro-structure noise and inefficiency. An adaptive Bayesian Markov Chain Monte Carlo method is developed and employed for estimation and forecasting, whose properties are assessed and compared with maximum likelihood, via a simulation study. Compared to a range of well-known parametric GARCH, GARCH with two-sided Weibull distribution and realized GARCH models, tail risk forecasting results across 7 market index return series and 2 individual assets clearly favor the realized GARCH models incorporating two-sided Weibull distribution, especially models employing the sub-sampled realized variance and sub-sampled realized range, over a six year period that includes the global financial crisis.

q-fin.RM↗

arXiv2017

Compressive fluorescence spectral imaging with a spectrometer

We present an efficient approach and principle experiment for compressive sensing (CS) fluorescence spectral imaging. According to the dimension-reduced effect of CS, the spectral and spatial information was simultaneously obtained by using a fiber spectrometer without mechanical scanning. As a method verification, we demonstrated spectral imaging under only two typical wavelengths, but the spectral resolution is up to 1.4nm depended on the fiber spectrometer. The method could obtain 50% light energy from the object, much larger compared with mechanical scanning which detects light of only one point per measurement. The relationship between sampling rate and image quality is also discussed in this study.

physics.ins-det↗

arXiv2017

Checking Linearizability of Concurrent Priority Queues

Efficient implementations of concurrent objects such as atomic collections are essential to modern computing. Programming such objects is error prone: in minimizing the synchronization overhead between concurrent object invocations, one risks the conformance to sequential specifications -- or in formal terms, one risks violating linearizability. Unfortunately, verifying linearizability is undecidable in general, even on classes of implementations where the usual control-state reachability is decidable. In this work we consider concurrent priority queues which are fundamental to many multi-threaded applications such as task scheduling or discrete event simulation, and show that verifying linearizability of such implementations can be reduced to control-state reachability. This reduction entails the first decidability results for verifying concurrent priority queues in the context of an unbounded number of threads, and it enables the application of existing safety-verification tools for establishing their correctness.

cs.PL↗

arXiv2017

Proof-of-principle experimental realization of a qubit-like qudit-based quantum key distribution scheme

In comparison to qubit-based protocols, qudit-based quantum key distribution (QKD) ones gen- erally allow two cooperative parties to share unconditionally secure keys under a higher channel noise. However, it is very hard to prepare and measure the required quantum states in qudit-based protocols in general. One exception is the recently proposed highly error tolerant qudit-based proto- col known as the Chau15 [1]. Remarkably, the state preparation and measurement in this protocol can be done relatively easily since the required states are phase encoded almost like the diagonal basis states of a qubit. Here we report the first proof-of-principle demonstration of the Chau15 protocol. One highlight of our experiment is that its post-processing is based on practical one-way manner, while the original proposal in Ref. [1] relies on complicated two-way post-processing, which is a great challenge in experiment. In addition, by manipulating time-bin qudit and measurement with a variable delay interferometer, our realization is extensible to qudit with high-dimensionality and confirms the experimental feasibility of the Chau15 protocol.

quant-ph↗

arXiv2017

Distributed Simulation Platform for Autonomous Driving

Autonomous vehicle safety and reliability are the paramount requirements when developing autonomous vehicles. These requirements are guaranteed by massive functional and performance tests. Conducting these tests on real vehicles is extremely expensive and time consuming, and thus it is imperative to develop a simulation platform to perform these tasks. For simulation, we can utilize the Robot Operating System (ROS) for data playback to test newly developed algorithms. However, due to the massive amount of simulation data, performing simulation on single machines is not practical. Hence, a high-performance distributed simulation platform is a critical piece in autonomous driving development. In this paper we present our experiences of building a production distributed autonomous driving simulation platform. This platform is built upon Spark distributed framework, for distributed computing management, and ROS, for data playback simulations.

cs.DC↗

arXiv2017

Spontaneous symmetry breaking induced unidirectional rotation of chain-grafted colloid in the active bath

Exploiting the energy of randomly moving active agents such as bacteria is a fascinating way to power a microdevice. Here we show, by simulations, that a chain-grafted disk-like colloid can rotate unidirectionally when immersed in a thin film of active particle suspension. The spontaneous symmetry breaking of chain configurations is the origin of the unidirectional rotation. Long persistence time, large propelling force and/or small rotating friction are keys to keeping the broken symmetry and realizing the rotation. In the rotating state, we find very simple linear relations, e.g. between mean angular speed and propelling force. The time-evolving asymmetry of chain configurations reveals that there are two types of non-rotating state. Our findings provide new insights into the phenomena of spontaneous symmetry breaking in active systems with flexible objects and also open the way to conceive new soft/deformable microdevices.

cond-mat.soft↗

arXiv2017

Self-assembled chiral phosphorus nanotubes from phosphorene: a molecular dynamics study

Controlled syntheses in nanoscale structures should be expected and phosphorous nanotubes with predefined chiralities are important in electronic devices with tunable bandgap. Here, incorporating molecular dynamics simulations with theoretical analyses, we show that a zigzag phosphorene nanoribbon can self-assemble and form a corresponding chiral phosphorous nanotube surrounding a template armchair phosphorous nanotube. The van der Waals potential between the nanoribbon and the nanotube is transformed to the intrinsic deformed and chemical bonding energies of the synthesized tube together with partial kinetic energy. The self-assembly process has an apparent temperature dependence and size effect and the formed chiral tube is thermodynamically stable. Also, the chirality and measurement can be tuned by the radius of template tube and the aspect ratio of raw ribbon. The study suggests a novel and feasible approach for controlled synthesis of phosphorous nanotubes and thus is of great interest for semiconductor device applications.

cond-mat.mtrl-sci↗

arXiv2017

Close Yet Distinctive Domain Adaptation

Domain adaptation is transfer learning which aims to generalize a learning model across training and testing data with different distributions. Most previous research tackle this problem in seeking a shared feature representation between source and target domains while reducing the mismatch of their data distributions. In this paper, we propose a close yet discriminative domain adaptation method, namely CDDA, which generates a latent feature representation with two interesting properties. First, the discrepancy between the source and target domain, measured in terms of both marginal and conditional probability distribution via Maximum Mean Discrepancy is minimized so as to attract two domains close to each other. More importantly, we also design a repulsive force term, which maximizes the distances between each label dependent sub-domain to all others so as to drag different class dependent sub-domains far away from each other and thereby increase the discriminative power of the adapted domain. Moreover, given the fact that the underlying data manifold could have complex geometric structure, we further propose the constraints of label smoothness and geometric structure consistency for label propagation. Extensive experiments are conducted on 36 cross-domain image classification tasks over four public datasets. The comprehensive results show that the proposed method consistently outperforms the state-of-the-art methods with significant margins.

cs.LG↗

arXiv2017

Implementing a Cloud Platform for Autonomous Driving

Autonomous driving clouds provide essential services to support autonomous vehicles. Today these services include but not limited to distributed simulation tests for new algorithm deployment, offline deep learning model training, and High-Definition (HD) map generation. These services require infrastructure support including distributed computing, distributed storage, as well as heterogeneous computing. In this paper, we present the details of how we implement a unified autonomous driving cloud infrastructure, and how we support these services on top of this infrastructure.

cs.DC↗

arXiv2017

Embedding compact surfaces into the 3-dimensional Euclidean space with maximum symmetry

The symmetries of surfaces which can be embedded into the symmetries of the 3-dimensional Euclidean space $\mathbb{R}^3$ are easier to feel by human's intuition. We give the maximum order of finite group actions on $(\mathbb{R}^3, Σ)$ among all possible embedded closed/bordered surfaces with given geometric/algebraic genus $>1$ in $\mathbb{R}^3$. We also identify the topological types of the bordered surfaces realizing the maximum order, and find simple representative embeddings for such surfaces.

math.GT↗

arXiv2017

Modified quantum-speed-limit bounds for open quantum dynamics in quantum channels

The minimal evolution time between two distinguishable states is of fundamental interest in quantum physics. Very recently Mirkin et al. argue that some most common quantum-speed-limit (QSL) bounds which depend on the actual evolution time do not cleave to the essence of the QSL theory as they grow indefinitely but the final state is reached at a finite time in a damped Jaynes-Cummings (JC) model. In this paper, we thoroughly study this puzzling phenomenon. We find the inconsistent estimates will happen if and only if the limit of resolution of a calculation program is achieved, through which we propose that the nature of the inconsistency is not a violation to the essence of the QSL theory but an illusion caused by the finite precision in numerical simulations. We also present a generic method to overcome the inconsistent estimates and confirm its effectiveness in both amplitude-damping and phase-damping channels. Additionally, we show special cases which may restrict the QSL bound defined by "quantumness".

quant-ph↗

arXiv2017

Exotic Spin Phases in Two Dimensional Spin-orbit Coupled Models: Importance of Quantum Fluctuation Effects

We investigate the phase diagrams of the effective spin models derived from Fermi-Hubbard and Bose-Hubbard models with Rashba spin-orbit coupling, using string bond states, one of the quantum tensor network states methods. We focus on the role of quantum fluctuation effect in stabilizing the exotic spin phases in these models. For boson systems, and when the ratio between inter-particle and intra-particle interaction $λ> 1$, the out-of-plane ferromagnetic (FM) and antiferromagnetic (AFM) phases obtained from quantum simulations are the same to those obtained from classic model. However, the quantum order-by-disorder effect reduces the classical in-plane XY-FM and XY-vortex phases to the quantum X/Y-FM and X/Y-stripe phase when $λ< 1$. The spiral phase and skyrmion phase can be realized in the presence of quantum fluctuation. For the Fermi-Hubbard model, the quantum fluctuation energies are always important in the whole parameter regime. %, which are much larger than those of the Bose model. A general picture to understand the phase diagrams from symmetry point of view is also presented.

cond-mat.str-el↗

arXiv2017

Perception-based energy functions in seam-cutting

Image stitching is challenging in consumer-level photography, due to alignment difficulties in unconstrained shooting environment. Recent studies show that seam-cutting approaches can effectively relieve artifacts generated by local misalignment. Normally, seam-cutting is described in terms of energy minimization, however, few of existing methods consider human perception in their energy functions, which sometimes causes that a seam with minimum energy is not most invisible in the overlapping region. In this paper, we propose a novel perception-based energy function in the seam-cutting framework, which considers the nonlinearity and the nonuniformity of human perception in energy minimization. Our perception-based approach adopts a sigmoid metric to characterize the perception of color discrimination, and a saliency weight to simulate that human eyes incline to pay more attention to salient objects. In addition, our seam-cutting composition can be easily implemented into other stitching pipelines. Experiments show that our method outperforms the seam-cutting method of the normal energy function, and a user study demonstrates that our composed results are more consistent with human perception.

cs.CV↗

arXiv2017

Compositional Model based Fisher Vector Coding for Image Classification

Deriving from the gradient vector of a generative model of local features, Fisher vector coding (FVC) has been identified as an effective coding method for image classification. Most, if not all, FVC implementations employ the Gaussian mixture model (GMM) to depict the generation process of local features. However, the representative power of the GMM could be limited because it essentially assumes that local features can be characterized by a fixed number of feature prototypes and the number of prototypes is usually small in FVC. To handle this limitation, in this paper we break the convention which assumes that a local feature is drawn from one of few Gaussian distributions. Instead, we adopt a compositional mechanism which assumes that a local feature is drawn from a Gaussian distribution whose mean vector is composed as the linear combination of multiple key components and the combination weight is a latent random variable. In this way, we can greatly enhance the representative power of the generative model of FVC. To implement our idea, we designed two particular generative models with such a compositional mechanism.

cs.CV↗