• In this paper we consider a general matrix factorization model which covers a large class of existing models with many applications in areas such as machine learning and imaging sciences. To solve this possibly nonconvex, nonsmooth and non-Lipschitz problem, we develop a non-monotone alternating updating method based on a potential function. Our method essentially updates two blocks of variables in turn by inexactly minimizing this potential function, and updates another auxiliary block of variables using an explicit formula. The special structure of our potential function allows us to take advantage of efficient computational strategies for non-negative matrix factorization to perform the alternating minimization over the two blocks of variables. A suitable line search criterion is also incorporated to improve the numerical performance. Under some mild conditions, we show that the line search criterion is well defined, and establish that the sequence generated is bounded and any cluster point of the sequence is a stationary point. Finally, we conduct some numerical experiments using real datasets to compare our method with some existing efficient methods for non-negative matrix factorization and matrix completion. The numerical results show that our method can outperform these methods for these specific applications.
  • In this letter, we propose a reliability-based windowed decoding scheme for spatially-coupled (SC) low-density parity-check (LDPC) codes. To mitigate the error propagation along the sliding windowed decoder of the SC LDPC codes, a partial message reservation (PMR) method is proposed where only the reliable messages generated in the previous decoding window are reserved for the next decoding window. We also propose a partial syndrome check (PSC) stopping rule for each decoding window, in which only the complete VNs are checked. Simulation results show that our proposed scheme significantly improves the error floor performance compared to the sliding windowed decoder with the conventional weighted bit-flipping (WBF) algorithm.
  • In a recent study, using the distribution of galaxies in the north galactic pole of SDSS DR7 region enclosed in a 500$\mpch$ box, we carried out our ELUCID simulation (Wang et al. 2016, ELUCID III). Here we {\it light} the dark matter halos and subhalos in the reconstructed region in the simulation with galaxies in the SDSS observations using a novel {\it neighborhood} abundance matching method. Before we make use of thus established galaxy-subhalo connections in the ELUCID simulation to evaluate galaxy formation models, we set out to explore the reliability of such a link. For this purpose, we focus on the following a few aspects of galaxies: (1) the central-subhalo luminosity and mass relations; (2) the satellite fraction of galaxies; (3) the conditional luminosity function (CLF) and conditional stellar mass function (CSMF) of galaxies; and (4) the cross correlation functions between galaxies and the dark matter particles, most of which are measured separately for all, red and blue galaxy populations. We find that our neighborhood abundance matching method accurately reproduces the central-subhalo relations, satellite fraction, the CLFs and CSMFs and the biases of galaxies. These features ensure that thus established galaxy-subhalo connections will be very useful in constraining galaxy formation processes. And we provide some suggestions on the three levels of using the galaxy-subhalo pairs for galaxy formation constraints. The galaxy-subhalo links and the subhalo merger trees in the SDSS DR7 region extracted from our ELUCID simulation are available upon request.
  • For the task of subdecimeter aerial imagery segmentation, the fine-grained semantic segmentation results are usually difficult to obtain because of complex remote sensing contents and optical conditions. In addition, remote sensing imagery has inherent limitations of imbalanced class distribution. Recently, convolutional neural networks (CNNs) have shown outstanding performance on this task. In this paper, we propose the TreeSegNet to solve the class imbalance problem and further improve the accuracy in the metrics' point of view. Based on the infrastructure of DeepUNet, a Tree-CNN model in which each node represents a ResNeXt unit is constructed automatically according to confusion matrix and minimum graph cut algorithm. By transporting feature maps by concatenating connections, the Tree-CNN block fuses the multiscale features and learning the best weights for the model. In the experiments on ISPRS 2D semantic labeling Potsdam dataset, the results gotten by TreeSegNet are better than the opened state-of-the-art methods. The F1 measure scores of classes are improved especially for those classes that are easily confused. Completely and detailed comparison and analysis are performed to show that the improvement is brought by the construction and the embedding of the Tree-CNN module.
  • Massive classification, a classification task defined over a vast number of classes (hundreds of thousands or even millions), has become an essential part of many real-world systems, such as face recognition. Existing methods, including the deep networks that achieved remarkable success in recent years, were mostly devised for problems with a moderate number of classes. They would meet with substantial difficulties, e.g. excessive memory demand and computational cost, when applied to massive problems. We present a new method to tackle this problem. This method can efficiently and accurately identify a small number of "active classes" for each mini-batch, based on a set of dynamic class hierarchies constructed on the fly. We also develop an adaptive allocation scheme thereon, which leads to a better tradeoff between performance and cost. On several large-scale benchmarks, our method significantly reduces the training cost and memory demand, while maintaining competitive performance.
  • Most multi-dimensional (more than two dimensions) lattice partitions only form additive quotient groups and lack multiplication operations. This prevents us from constructing lattice codes based on multi-dimensional lattice partitions directly from non-binary linear codes over finite fields. In this paper, we design lattice codes from Construction A lattices where the underlying linear codes are non-binary irregular repeat-accumulate (IRA) codes. Most importantly, our codes are based on multi-dimensional lattice partitions with finite constellations. We propose a novel encoding structure that adds randomly generated lattice sequences to the encoder's messages, instead of multiplying lattice sequences to the encoder's messages. We prove that our approach can ensure that the decoder's messages exhibit permutation-invariance and symmetry properties. With these two properties, the densities of the messages in the iterative decoder can be modeled by Gaussian distributions described by a single parameter. With Gaussian approximation, extrinsic information transfer (EXIT) charts for our multi-dimensional IRA lattice codes are developed and used for analyzing the convergence behavior and optimizing the decoding thresholds. Simulation results show that our codes can approach the unrestricted Shannon limit within 0.46 dB and outperform the previously designed lattice codes with two-dimensional lattice partitions and existing lattice coding schemes for large codeword length.
  • We propose a new class of information-coupled (IC) Turbo codes to improve the transport block (TB) error rate performance for long-term evolution (LTE) systems, while keeping the hybrid automatic repeat request protocol and the Turbo decoder for each code block (CB) unchanged. In the proposed codes, every two consecutive CBs in a TB are coupled together by sharing a few common information bits. We propose a feed-forward and feed-back decoding scheme and a windowed (WD) decoding scheme for decoding the whole TB by exploiting the coupled information between CBs. Both decoding schemes achieve a considerable signal-to-noise-ratio (SNR) gain compared to the LTE Turbo codes. We construct the extrinsic information transfer (EXIT) functions for the LTE Turbo codes and our proposed IC Turbo codes from the EXIT functions of underlying convolutional codes. An SNR gain upper bound of our proposed codes over the LTE Turbo codes is derived and calculated by the constructed EXIT charts. Numerical results show that the proposed codes achieve an SNR gain of 0.25 dB to 0.72 dB for various code parameters at a TB error rate level of $10^{-2}$, which complies with the derived SNR gain upper bound.
  • Halo bias is the one of the key ingredients of the halo models. It was shown at a given redshift to be only dependent, to the first order, on the halo mass. In this study, four types of cosmic web environments: clusters, filaments, sheets and voids are defined within a state of the art high resolution $N$-body simulation. Within those environments, we use both halo-dark matter cross-correlation and halo-halo auto correlation functions to probe the clustering properties of halos. The nature of the halo bias differs strongly among the four different cosmic web environments we describe. With respect to the overall population, halos in clusters have significantly lower biases in the {$10^{11.0}\sim 10^{13.5}h^{-1}\rm M_\odot$} mass range. In other environments however, halos show extremely enhanced biases up to a factor 10 in voids for halos of mass {$\sim 10^{12.0}h^{-1}\rm M_\odot$}. Such a strong cosmic web environment dependence in the halo bias may play an important role in future cosmological and galaxy formation studies. Within this cosmic web framework, the age dependency of halo bias is found to be only significant in clusters and filaments for relatively small halos $\la 10^{12.5}\msunh$.
  • Semantic segmentation is a fundamental research in remote sensing image processing. Because of the complex maritime environment, the sea-land segmentation is a challenging task. Although the neural network has achieved excellent performance in semantic segmentation in the last years, there are a few of works using CNN for sea-land segmentation and the results could be further improved. This paper proposes a novel deep convolution neural network named DeepUNet. Like the U-Net, its structure has a contracting path and an expansive path to get high resolution output. But differently, the DeepUNet uses DownBlocks instead of convolution layers in the contracting path and uses UpBlock in the expansive path. The two novel blocks bring two new connections that are U-connection and Plus connection. They are promoted to get more precise segmentation results. To verify our network architecture, we made a new challenging sea-land dataset and compare the DeepUNet on it with the SegNet and the U-Net. Experimental results show that DeepUNet achieved good performance compared with other architectures, especially in high-resolution remote sensing imagery.
  • For vertical velocity field $v_{\rm z} (r,z;R)$ of granular flow through an aperture of radius $R$, we propose a size scaling form $v_{\rm z}(r,z;R)=v_{\rm z} (0,0;R)f (r/R_{\rm r}, z/R_{\rm z})$ in the region above the aperture. The length scales $R_{\rm r}=R- 0.5 d$ and $R_{\rm z}=R+k_2 d$, where $k_2$ is a parameter to be determined and $d$ is the diameter of granule. The effective acceleration, which is derived from $v_{\rm z}$, follows also a size scaling form $a_{\rm eff} = v_{\rm z}^2(0,0;R)R_{\rm z}^{-1} \theta (r/R_{\rm r}, z/R_{\rm z})$. For granular flow under gravity $g$, there is a boundary condition $a_{\rm eff} (0,0;R)=-g$ which gives rise to $v_{\rm z} (0,0;R)= \sqrt{ \lambda g R_{\rm z}}$ with $\lambda=-1/\theta (0,0)$. Using the size scaling form of vertical velocity field and its boundary condition, we can obtain the flow rate $W =C_2 \rho \sqrt{g } R_{\rm r}^{D-1} R_{\rm z}^{1/2} $, which agrees with the Beverloo law when $R \gg d$. The vertical velocity fields $v_z (r,z;R)$ in three-dimensional (3D) and two-dimensional (2D) hoppers have been simulated using the discrete element method (DEM) and GPU program. Simulation data confirm the size scaling form of $v_{\rm z} (r,z;R)$ and the $R$-dependence of $v_{\rm z} (0,0;R)$.
  • Fast radio bursts are radio transients observed mainly around 1.5 GHz. Their peak frequency decreases at a rate of 100 ~ 500 MHz/s and some of them have a broader pulse with an exponentially decaying tail. Common assumptions for fast radio bursts include a dispersion effect resulting in the peak frequency drifting and a scattering effect resulting in pulse broadening. These assumptions attribute the abnormally large dispersion measure and scattering measure to the environmental medium of the host galaxy. Here we show that the radiation of fast radio bursts can be explained as an undulator radiation and the large dispersion measure can be due to a motion effect mainly from the rotation of the source which is probably variable stars. In our scenario, the pulse broadening is near-field effects and the pulse itself represents a Fresnel diffraction pattern sweeping the observer. Our work is the first analysis of properties of fast radio bursts in the context of a special mechanism of the radiation instead of a special propagation environment of the radiation.
  • Deployment of billions of Commercial off-the-shelf (COTS) RFID tags has drawn much of the attention of the research community because of the performance gaps of current systems. In particular, hash-enabled protocol (HEP) is one of the most thoroughly studied topics in the past decade. HEPs are designed for a wide spectrum of notable applications (e.g., missing detection) without need to collect all tags. HEPs assume that each tag contains a hash function, such that a tag can select a random but predicable time slot to reply with a one-bit presence signal that shows its existence. However, the hash function has never been implemented in COTS tags in reality, which makes HEPs a 10-year untouchable mirage. This work designs and implements a group of analog on-tag hash primitives (called Tash) for COTS Gen2-compatible RFID systems, which moves prior HEPs forward from theory to practice. In particular, we design three types of hash primitives, namely, tash function, tash table function and tash operator. All of these hash primitives are implemented through selective reading, which is a fundamental and mandatory functionality specified in Gen2 protocol, without any hardware modification and fabrication. We further apply our hash primitives in two typical HEP applications (i.e., cardinality estimation and missing detection) to show the feasibility and effectiveness of Tash. Results from our prototype, which is composed of one ImpinJ reader and 3, 000 Alien tags, demonstrate that the new design lowers 60% of the communication overhead in the air. The tash operator can additionally introduce an overhead drop of 29.7%.
  • Air traffic is widely known as a complex, task-critical techno-social system, with numerous interactions between airspace, procedures, aircraft and air traffic controllers. In order to develop and deploy high-level operational concepts and automation systems scientifically and effectively, it is essential to conduct an in-depth investigation on the intrinsic traffic-human dynamics and characteristics, which is not widely seen in the literature. To fill this gap, we propose a multi-layer network to model and analyze air traffic systems. A Route-based Airspace Network (RAN) and Flight Trajectory Network (FTN) encapsulate critical physical and operational characteristics; an Integrated Flow-Driven Network (IFDN) and Interrelated Conflict-Communication Network (ICCN) are formulated to represent air traffic flow transmissions and intervention from air traffic controllers, respectively. Furthermore, a set of analytical metrics including network variables, complex network attributes, controllers' cognitive complexity, and chaotic metrics are introduced and applied in a case study of Guangzhou terminal airspace. Empirical results show the existence of fundamental diagram and macroscopic fundamental diagram at the route, sector and terminal levels. Moreover, the dynamics and underlying mechanisms of "ATCOs-flow" interactions are revealed and interpreted by adaptive meta-cognition strategies based on network analysis of the ICCN. Finally, at the system level, chaos is identified in conflict system and human behavioral system when traffic switch to the semi-stable or congested phase. This study offers analytical tools for understanding the complex human-flow interactions at potentially a broad range of air traffic systems, and underpins future developments and automation of intelligent air traffic management systems.
  • In many applications involving large dataset or online updating, stochastic gradient descent (SGD) provides a scalable way to compute parameter estimates and has gained increasing popularity due to its numerical convenience and memory efficiency. While the asymptotic properties of SGD-based estimators have been established decades ago, statistical inference such as interval estimation remains much unexplored. The traditional resampling method such as the bootstrap is not computationally feasible since it requires to repeatedly draw independent samples from the entire dataset. The plug-in method is not applicable when there are no explicit formulas for the covariance matrix of the estimator. In this paper, we propose a scalable inferential procedure for stochastic gradient descent, which, upon the arrival of each observation, updates the SGD estimate as well as a large number of randomly perturbed SGD estimates. The proposed method is easy to implement in practice. We establish its theoretical properties for a general class of models that includes generalized linear models and quantile regression models as special cases. The finite-sample performance and numerical utility is evaluated by simulation studies and two real data applications.
  • The $E_8$ root lattice can be constructed from the modular curve $X(13)$ by the invariant theory for the simple group $\text{PSL}(2, 13)$. This gives a different construction of the $E_8$ root lattice. It also gives an explicit construction of the modular curve $X(13)$.
  • Given a homogeneous space $X = G/\Gamma$ with $G$ containing the group $H = (\mathrm{SO}(n,1))^k$. Let $x\in X$ such that $Hx$ is dense in $X$. Given an analytic curve $\phi: I=[a,b] \rightarrow H$, we will show that if $\phi$ satisfies certain geometric condition, then for a typical diagonal subgroup $A =\{a(t): t \in \mathbb{R}\} \subset H$ the translates $\{a(t)\phi(I)x: t >0\}$ of the curve $\phi(I)x$ will tend to be equidistributed in $X$ as $t \rightarrow +\infty$. The proof is based on the study of linear representations of $\mathrm{SO}(n,1)$ and $H$.
  • In this paper, we study the weighted $n$-dimensional badly approximable points on manifolds. Given a $C^n$ differentiable non-degenerate submanifold $\mathcal{U} \subset \mathbb{R}^n$, we will show that any countable intersection of the sets of the weighted badly approximable points on $\mathcal{U}$ has full Hausdorff dimension. This strengthens a result of Beresnevich by removing the condition on weights and weakening the smoothness condition on manifolds. Compared to the work of Beresnevich, our approach relies on homogeneous dynamics. It turns out that in order to solve this problem, it is crucial to study the distribution of long pieces of unipotent orbits in homogeneous spaces. The proof relies on the linearization technique and representations of $\mathrm{SL}(n+1,\mathbb{R})$.
  • The integration of cloud computing and Internet of Things (IoT) is quickly becoming the key enabler for the digital transformation of the healthcare industry by offering comprehensive improvements in patient engagements, productivity and risk mitigation. This paradigm shift, while bringing numerous benefits and new opportunities to healthcare organizations, has raised a lot of security and privacy concerns. In this paper, we present a reliable, searchable and privacy-preserving e-healthcare system, which takes advantage of emerging cloud storage and IoT infrastructure and enables healthcare service providers (HSPs) to realize remote patient monitoring in a secure and regulatory compliant manner. Our system is built upon a novel dynamic searchable symmetric encryption scheme with forward privacy and delegated verifiability for periodically generated healthcare data. While the forward privacy is achieved by maintaining an increasing counter for each keyword at an IoT gateway, the data owner delegated verifiability comes from the combination of the Bloom filter and aggregate message authentication code. Moreover, our system is able to support multiple HSPs through either data owner assistance or delegation. The detailed security analysis as well as the extensive simulations on a large data set with millions of records demonstrate the practical efficiency of the proposed system for real world healthcare applications.
  • Using a method to correct redshift space distortion (RSD) for individual galaxies, we mapped the real space distributions of galaxies in the Sloan Digital Sky Survey (SDSS) Data Release 7 (DR7). We use an ensemble of mock catalogs to demonstrate the reliability of our method. Here as the first paper in a series, we mainly focus on the two point correlation function (2PCF) of galaxies. Overall the 2PCF measured in the reconstructed real space for galaxies brighter than $^{0.1}{\rm M}_r-5\log h=-19.0$ agrees with the direct measurement to an accuracy better than the measurement error due to cosmic variance, if the reconstruction uses the correct cosmology. Applying the method to the SDSS DR7, we construct a real space version of the main galaxy catalog, which contains 396,068 galaxies in the North Galactic Cap with redshifts in the range $0.01 \leq z \leq 0.12$. The Sloan Great Wall, the largest known structure in the nearby Universe, is not as dominant an over-dense structure as appears to be in redshift space. We measure the 2PCFs in reconstructed real space for galaxies of different luminosities and colors. All of them show clear deviations from single power-law forms, and reveal clear transitions from 1-halo to 2-halo terms. A comparison with the corresponding 2PCFs in redshift space nicely demonstrates how RSDs boost the clustering power on large scales (by about $40-50\%$ at scales $\sim 10 h^{-1}{\rm {Mpc}}$) and suppress it on small scales (by about $70-80\%$ at a scale of $0.3 h^{-1}{\rm {Mpc}}$).
  • In this paper, we study a general optimization model, which covers a large class of existing models for many applications in imaging sciences. To solve the resulting possibly nonconvex, nonsmooth and non-Lipschitz optimization problem, we adapt the alternating direction method of multipliers (ADMM) with a general dual step-size to solve a reformulation that contains three blocks of variables, and analyze its convergence. We show that for any dual step-size less than the golden ratio, there exists a computable threshold such that if the penalty parameter is chosen above such a threshold and the sequence thus generated by our ADMM is bounded, then the cluster point of the sequence gives a stationary point of the nonconvex optimization problem. We achieve this via a potential function specifically constructed for our ADMM. Moreover, we establish the global convergence of the whole sequence if, in addition, this special potential function is a Kurdyka-{\L}ojasiewicz function. Furthermore, we present a simple strategy for initializing the algorithm to guarantee boundedness of the sequence. Finally, we perform numerical experiments comparing our ADMM with the proximal alternating linearized minimization (PALM) proposed in [5] on the background/foreground extraction problem with real data. The numerical results show that our ADMM with a nontrivial dual step-size is efficient.
  • Let $H = \mathrm{SO}(n,1)$ and $A = \{a(t): t \in \mathbb{R}\}$ be a maximal $\mathbb{R}$-split Cartan subgroup of $H$. Let $\Gamma \subset H \times H$ be a nonuniform lattice in $H \times H$ and $X_{\Gamma} : = H \times H/ \Gamma$. Let $A_2 : = \{ a_2(t):=a(t) \times a(t) : t \in \mathbb{R}\} \subset A\times A$ on $X_{\Gamma}$ and $\mathcal{D}_{\Gamma}\subset X_{\Gamma}$ denote the collection of points $x \in X_{\Gamma}$ such that $a_2(t)x$ diverges as $t \rightarrow +\infty$. In this note, we will show that if the Hausdorff dimension of $\mathcal{D}_{\Gamma}$ is greater than $\dim (H\times H) - 2(n-1)$, then $\Gamma $ is essentially split, namely, it contains a subgroup of finite index of form $\Gamma_1 \times \Gamma_2 $, where $\Gamma_1$ and $\Gamma_2$ are both lattices in $H$.
  • We show that genuinely higher rank expanding actions of abelian semi-groups on compact manifolds are $C^{\infty}$-conjugate to affine actions on infra-nilmanifolds. This is based on the classification of expanding diffeomorphisms up to \holder conjugacy by Gromov and Shub, and is similar to recent work on smooth classification of higher rank Anosov actions on tori and nilmanifolds. To prove regularity of the conjugacy in the higher rank setting, we establish exponential mixing of solenoid actions induced from semi-group actions by nilmanifold endomorphisms, a result of independent interest. We then proceed similar to the case of higher rank Anosov actions.
  • Rock-salt structured chalcogenides, such as PbTe, PbSe, and SnTe, are the top candidates for mid-temperature thermoelectric applications, and their p-type thermoelectric efficiencies can be enhanced via aligning the valence bands. Here, we provided comprehensive numerical investigations on the effects of band convergence on electronic properties. We found that the extra valance band can indeed significantly enhance the power factor. Nevertheless, the extra valance band can also increase the electronic thermal conductivity, which partially offsets the enhanced power factor for the overall figure-of-merit. Finally, we predicted that the maximum figure-of-merit for PbTe, PbSe, and SnTe can reach to 2.2, 1.8, and 1.6, re-spectively, without relying on the reduction in lattice thermal conductivity.
  • Distributed spectrum access (DSA) is challenging since an individual secondary user often has limited sensing capabilities only. One key insight is that channel recommendation among secondary users can help to take advantage of the inherent correlation structure of spectrum availability in both time and space, and enable users to obtain more informed spectrum opportunities. With this insight, we advocate to leverage the wisdom of crowds, and devise social recommendation aided DSA mechanisms to orient secondary users to make more intelligent spectrum access decisions, for both strong and weak network information cases. We start with the strong network information case where secondary users have the statistical information. To mitigate the difficulty due to the curse of dimensionality in the stochastic game approach, we take the one-step Nash approach and cast the social recommendation aided DSA decision making problem at each time slot as a strategic game. We show that it is a potential game, and then devise an algorithm to achieve the Nash equilibrium by exploiting its finite improvement property. For the weak information case where secondary users do not have the statistical information, we develop a distributed reinforcement learning mechanism for social recommendation aided DSA based on the local observations of secondary users only. Appealing to the maximum-norm contraction mapping, we also derive the conditions under which the distributed mechanism converges and characterize the equilibrium therein. Numerical results reveal that the proposed social recommendation aided DSA mechanisms can achieve superior performance using real social data traces and its performance loss in the weak network information case is insignificant, compared with the strong network information case.
  • In this paper, we study an analytic curve $\varphi: I=[a,b]\rightarrow \mathrm{M}(m\times n, \mathbb{R})$ in the space of $m$ by $n$ real matrices, and show that if $\varphi$ satisfies certain geometric condition, then for almost every point on the curve, the Diophantine approximation given by Dirichlet's Theorem can not be improved. To do this, we embed the curve into some homogeneous space $G/\Gamma$, and prove that under the action of some expanding diagonal subgroup $A= \{a(t): t \in \mathbb{R}\}$, the translates of the curve tend to be equidistributed in $G/\Gamma$, as $t \rightarrow +\infty$.