arXiv subjects

Bo Yang

Publications and source records attributed to Bo Yang.

At least 559 records · Page 31Linked to original sources

arXiv2016

Robust Volume Minimization-Based Matrix Factorization for Remote Sensing and Document Clustering

This paper considers \emph{volume minimization} (VolMin)-based structured matrix factorization (SMF). VolMin is a factorization criterion that decomposes a given data matrix into a basis matrix times a structured coefficient matrix via finding the minimum-volume simplex that encloses all the columns of the data matrix. Recent work showed that VolMin guarantees the identifiability of the factor matrices under mild conditions that are realistic in a wide variety of applications. This paper focuses on both theoretical and practical aspects of VolMin. On the theory side, exact equivalence of two independently developed sufficient conditions for VolMin identifiability is proven here, thereby providing a more comprehensive understanding of this aspect of VolMin. On the algorithm side, computational complexity and sensitivity to outliers are two key challenges associated with real-world applications of VolMin. These are addressed here via a new VolMin algorithm that handles volume regularization in a computationally simple way, and automatically detects and {iteratively downweights} outliers, simultaneously. Simulations and real-data experiments using a remotely sensed hyperspectral image and the Reuters document corpus are employed to showcase the effectiveness of the proposed algorithm.

stat.ML↗

arXiv2016

On Bismut Flat Manifolds

In this paper, we give a classification of all compact Hermitian manifolds with flat Bismut connection. We show that the torsion tensor of such a manifold must be parallel, thus the universal cover of such a manifold is a Lie group equipped with a bi-invariant metric and a compatible left invariant complex structure. In particular, isosceles Hopf surfaces are the only Bismut flat compact non-Kähler surfaces, while central Calabi-Eckmann threefolds are the only simply-connected compact Bismut flat threefolds.

math.DG↗

arXiv2016

Parallel Triangular Solvers on GPU

In this paper, we investigate GPU based parallel triangular solvers systematically. The parallel triangular solvers are fundamental to incomplete LU factorization family preconditioners and algebraic multigrid solvers. We develop a new matrix format suitable for GPU devices. Parallel lower triangular solvers and upper triangular solvers are developed for this new data structure. With these solvers, ILU preconditioners and domain decomposition preconditioners are developed. Numerical results show that we can speed triangular solvers around seven times faster.

cs.MS↗

arXiv2016

Development of Krylov and AMG linear solvers for large-scale sparse matrices on GPUs

This research introduce our work on developing Krylov subspace and AMG solvers on NVIDIA GPUs. As SpMV is a crucial part for these iterative methods, SpMV algorithms for single GPU and multiple GPUs are implemented. A HEC matrix format and a communication mechanism are established. And also, a set of specific algorithms for solving preconditioned systems in parallel environments are designed, including ILU(k), RAS and parallel triangular solvers. Based on these work, several Krylov solvers and AMG solvers are developed. According to numerical experiments, favorable acceleration performance is acquired from our Krylov solver and AMG solver under various parameter conditions.

cs.MS↗

arXiv2016

Learning From Hidden Traits: Joint Factor Analysis and Latent Clustering

Dimensionality reduction techniques play an essential role in data analytics, signal processing and machine learning. Dimensionality reduction is usually performed in a preprocessing stage that is separate from subsequent data analysis, such as clustering or classification. Finding reduced-dimension representations that are well-suited for the intended task is more appealing. This paper proposes a joint factor analysis and latent clustering framework, which aims at learning cluster-aware low-dimensional representations of matrix and tensor data. The proposed approach leverages matrix and tensor factorization models that produce essentially unique latent representations of the data to unravel latent cluster structure -- which is otherwise obscured because of the freedom to apply an oblique transformation in latent space. At the same time, latent cluster structure is used as prior information to enhance the performance of factorization. Specific contributions include several custom-built problem formulations, corresponding algorithms, and discussion of associated convergence properties. Besides extensive simulations, real-world datasets such as Reuters document data and MNIST image data are also employed to showcase the effectiveness of the proposed approaches.

cs.LG↗

arXiv2016

Cluster Statistics and Quasisoliton Dynamics in Microscopic Car-following Models

Using the optimal velocity (OV) model as an example, we show that in the non-linear regime there is an emergent quantity that gives the extremum headways in the cluster formation, as well as the coexistence curve separating the absolute stable phase from the metastable phase. This emergent quantity is independent of the density of the traffic lane, and determines an intrinsic scale that characterizes the dynamics of localized quasisoliton structures given by the time derivative of the headways. The intrinsic scale is analogous to the "charge" of quasisolitons that controls the strength of interaction between multiple clusters, leading to non-trivial cluster statistics from random perturbations to initial uniform traffic. The cluster statistics depend both on the charge and the density of the traffic lane; the relationship is qualitatively universal for general car-following models.

cond-mat.stat-mech↗

arXiv2016

The set of all orthogonal complex structures on the flat $6$-tori

In \cite{BSV}, Borisov, Salamon and Viaclovsky constructed non-standard orthogonal complex structures on flat tori $T^{2n}_{\mathbb R}$ for any $n\geq 3$. We will call these examples BSV-tori. In this note, we show that on a flat $6$-torus, all the orthogonal complex structures are either the complex tori or the BSV-tori. This solves the classification problem for compact Hermitian manifolds with flat Riemannian connection in the case of complex dimension three.

math.DG↗

arXiv2016

Energy Efficient Resource Allocation for Time-Varying OFDMA Relay Systems with Hybrid Energy Supplies

This paper investigates the energy efficient resource allocation for orthogonal frequency division multiple access (OFDMA) relay systems, where the system is supplied by the conventional utility grid and a renewable energy generator equipped with a storage device. The optimal usage of radio resource depends on the characteristics of the renewable energy generation and the mobile traffic, which exhibit both temporal and spatial diversities. Lyapunov optimization method is used to decompose the problem into the joint flow control, radio resource allocation and energy management without knowing a priori knowledge of system statistics. It is proven that the proposed algorithm can result in close-to-optimal performance with capacity limited data buffer and storage device. Simulation results show that the flexible tradeoff between the system utility and the conventional energy consumption can be achieved. Compared with other schemes, the proposed algorithm demonstrates better performance.

cs.DC↗

arXiv2016

Microscopic Statistical Characterisation of the Congested Traffic Flow and Some Salient Empirical Features

We present large scale and detailed analysis of the microscopic empirical data of the traffic flow, focusing on the non-linear interactions between the vehicles when the traffic is congested. By implementing a "renormalisation" procedure that averages over relatively unimportant factors, we extract the effective dependence of the acceleration on the vehicle headway, velocity and relative velocity, that characterises not just a few vehicles but the traffic system as a whole. Several interesting features of the detailed vehicle-to-vehicle interactions are revealed, including the stochastic distribution of human responses, relative importance of non-linear terms in different density regimes, symmetric response to the relative velocity, and the insensitivity of the acceleration to the velocity within a certain headway and velocity range. The latter leads to a multitude of steady-states without a fundamental diagram, showing strong evidence of the "synchronised phase", substantiated with additional microscopic details. We discuss the richness of the available data for the understanding of the human driving behaviours, and its usefulness in construction and calibration of both deterministic and stochastic microscopic traffic models.

nlin.AO↗

arXiv2016

Distributed Control for Charging Multiple Electric Vehicles with Overload Limitation

Severe pollution induced by traditional fossil fuels arouses great attention on the usage of plug-in electric vehicles (PEVs) and renewable energy. However, large-scale penetration of PEVs combined with other kinds of appliances tends to cause excessive or even disastrous burden on the power grid, especially during peak hours. This paper focuses on the scheduling of PEVs charging process among different charging stations and each station can be supplied by both renewable energy generators and a distribution network. The distribution network also powers some uncontrollable loads. In order to minimize the on-grid energy cost with local renewable energy and non-ideal storage while avoiding the overload risk of the distribution network, an online algorithm consisting of scheduling the charging of PEVs and energy management of charging stations is developed based on Lyapunov optimization and Lagrange dual decomposition techniques. The algorithm can satisfy the random charging requests from PEVs with provable performance. Simulation results with real data demonstrate that the proposed algorithm can decrease the time-average cost of stations while avoiding overload in the distribution network in the presence of random uncontrollable loads.

eess.SY↗

arXiv2016

Backhaul-Aware User Association and Resource Allocation for Energy-Constrained HetNets

Growing attentions have been paid to renewable energy or hybrid energy powered heterogeneous networks (HetNets). In this paper, focusing on backhaul-aware joint user association and resource allocation for this type of HetNets, we formulate an online optimization problem to maximize the network utility reflecting proportional fairness. Since user association and resource allocation are tightly coupled not only on resource consumption of the base stations (BSs), but also in the constraints of their available energy and backhaul, the closed-form solution is quite difficult to obtain. Thus, we solve the problem distributively via employing some decomposition methods. Specifically, at first, by adopting primal decomposition method, we decompose the original problem into a lower-level resource allocation problem for each BS, and a higher-level user association problem. For the optimal resource allocation, we prove that a BS either assigns equal normalized resources or provides equal long-term service rate to its served users. Then, the user association problem is solved by Lagrange dual decomposition method, and a completely distributed algorithm is developed. Moreover, applying results of the subgradient method, we demonstrate the convergence of the proposed distributed algorithm. Furthermore, in order to efficiently and reliably apply the proposed algorithm to the future wireless networks with an extremely dense BS deployment, we design a virtual user association and resource allocation scheme based on the software-defined networking architecture. Lastly, numerical results validate the convergence of the proposed algorithm and the significant improvement on network utility, load balancing and user fairness.

cs.NI↗

arXiv2015

Computationally Designed Zirconium Organometallic Catalyst for Direct Epoxidation of Alkenes without Allylic H Atoms: Aromatic Linkage Eliminates Formation of Inert Octahedral Complexes

We used density functional theory to computationally design a Zr organometallic catalyst for selectively oxidizing substrates using molecular oxygen as oxidant without coreductant. Each selective oxidation cycle involves four general steps: (a) a peroxo or weakly adsorbed O2 group releases an O atom to substrate to form substrate oxide and an oxo group, (b) an oxygen molecule adds to the oxo group to generate an eta2-ozone group, (c) the eta2-ozone group rearranges to form an eta3-ozone group, and (d) the eta3-ozone group releases an O atom to substrate to form substrate oxide and regenerate the peroxo or weakly adsorbed O2 group. This catalyst could potentially be synthesized via the condensation reaction Zr(N(R)R')4 + 2 C6H4-1,6-(N(C6H3-2',6'-(CH(CH3)2)2)OH)2 --> Zr(C6H4-1,6-(N(C6H3-2',6'-(CH(CH3)2)2)O)2)2 [aka Zr_Benzol catalyst] + 4 N(R)(R')H where R and R' are CH3, CH2CH3, or other alkyl groups. For direct ethylene epoxidation, the computed enthalpic energetic span (i.e., effective activation energy for the entire catalytic cycle) is 27.1 kcal/mol, which is one of the lowest values for catalysts studied to date. We study reaction mechanisms and the stability of different catalyst forms as a function of the oxygen atom chemical potential. Notably, an aromatic linkage in each ligand prevents this catalyst from deactivating to form an inactive octahedral-like structure that contains the same atoms as the dioxo complex, Zr(Ligand)2(O)2. Due to a side reaction that can transfer an allylic H atom from alkene to catalyst, this catalyst is useful for directly epoxidizing alkenes such as ethylene that do not contain allylic H atoms. To better understand the reaction chemistry, we computed net atomic charges and bond orders for the two catalytically relevant reaction cycles. These results quantify electron transfer and bond forming and breaking during the catalytic process.

physics.chem-ph↗

arXiv2015

Nonlinear Image Formation by Optical Superlattices

The angular spectrum theory is applied to the nonlinear harmonic generation process in optical superlattices. Several explicit and analytical structure functions are deduced to design optical superlattices for various purposes. Employing this method, nonlinear image formation is achieved during the second-harmonic generation process in a properly designed optical superlattice. This method is universal for both nonlinear beam shaping and nonlinear arbitrary image formation. The theory has been experimentally validated in two-dimensional optical superlattice of LiTaO3 crystals and the results agree well with the theoretical prediction. This work not only extends the application of optical superlattices, but also opens a new area for imaging technologies.

physics.optics↗

arXiv2015

On-surface synthesis of graphene nanoribbons with zigzag edge topology

Graphene-based nanostructures exhibit a vast range of exciting electronic properties that are absent in extended graphene. For example, quantum confinement in carbon nanotubes and armchair graphene nanoribbons (AGNRs) leads to the opening of substantial electronic band gaps that are directly linked to their structural boundary conditions. Even more intriguing are nanostructures with zigzag edges, which are expected to host spin-polarized electronic edge states and can thus serve as key elements for graphene-based spintronics. The most prominent example is zigzag graphene nanoribbons (ZGNRs) for which the edge states are predicted to couple ferromagnetically along the edge and antiferromagnetically between them. So far, a direct observation of the spin-polarized edge states for specifically designed and controlled zigzag edge topologies has not been achieved. This is mainly due to the limited precision of current top-down approaches, which results in poorly defined edge structures. Bottom-up fabrication approaches, on the other hand, were so far only successfully applied to the growth of AGNRs and related structures. Here, we describe the successful bottom-up synthesis of ZGNRs, which are fabricated by the surface-assisted colligation and cyclodehydrogenation of specifically designed precursor monomers including carbon groups that yield atomically precise zigzag edges. Using scanning tunnelling spectroscopy we prove the existence of edge-localized states with large energy splittings. We expect that the availability of ZGNRs will finally allow the characterization of their predicted spin-related properties such as spin confinement and filtering, and ultimately add the spin degree of freedom to graphene-based circuitry.

cond-mat.mtrl-sci↗

arXiv2015

Classification and Unification of the Microscopic Deterministic Traffic Models with Identical Drivers

We show that all existing deterministic microscopic traffic models with identical drivers (including both two-phase and three-phase models) can be understood as special cases from a master model by expansion around well-defined ground states. This allows two traffic models to be compared in a well-defined way. The three-phase models are characterized by the vanishing of leading orders of expansion within a certain density range, and as an example the popular intelligent driver models (IDM) is shown to be equivalent to a generalized optimal velocity (OV) model. We also explore the diverse solutions of the generalized OV model that can be important both for understanding human driving behaviors and algorithms for autonomous driverless vehicles.

nlin.AO↗

arXiv2015

Randomization Improving Online Time-Sensitive Revenue Maximization for Green Data Centers

Green data centers have become more and more popular recently due to their sustainability. The resource management module within a green data center, which is in charge of dispatching jobs and scheduling energy, becomes especially critical as it directly affects a center's profit and sustainability. The thrust of managing a green data center's machine and energy resources lies at the uncertainty of incoming job requests and future showing-up green energy supplies. Thus, the decision of scheduling resources has to be made in an online manner. Some heuristic deterministic online algorithms have been proposed in recent literature. In this paper, we consider online algorithms for green data centers and introduce a randomized solution with the objective of maximizing net profit. Competitive analysis is employed to measure online algorithms' theoretical performance. Our algorithm is theoretical-sound and it outperforms the previously known deterministic algorithms in many settings using real traces. To complement our study, optimal offline algorithms are also designed.

cs.DC↗

arXiv2015

Dynamic Sleep Control in Green Relay-Assisted Networks for Energy Saving and QoS Improving

We study the relay station (RS) sleep control mechanism targeting on reducing energy consumption while improving users' quality of service (QoS) in green relay-assisted cellular networks, where the base station (BS) is powered by grid power and the RSs are powered by renewable energy. By adopting green RSs, the grid power consumption of the BS is greatly reduced. But due to the uncertainty and stochastic characteristics of the renewable energy, power supply for RSs is not always sufficient. Thus the harvested energy needs to be scheduled appropriately to cater to the dynamic traffic so as to minimize the energy saving in the long term. An optimization problem is formulated to find the optimal sleep ratio of RSs to match the time variation of energy harvesting and traffic arrival. To fully use the renewable energy, green-RS-first principle is adopted in the user association process. The optimal RS sleeping policy is obtained through dynamic programming (DP) approach, which divides the original optimization problem into per-stage subproblems. A reduced DP algorithm and a greedy algorithm are further proposed to greatly reduce the computation complexity. By simulations, the reduced DP algorithm outperforms the greedy algorithm in achieving satisfactory energy saving and QoS performance.

cs.NI↗

arXiv2015

Dirac Cone Metric and the Origin of the Spin Connections in Monolayer Graphene

We show that the modulation of the hopping amplitudes in the honeycomb lattice of the monolayer graphene uniquely defines a metric which corresponds to the shape of the Dirac cone. The spin connection of this effective metric field can be obtained from the microscopic tight-binding Hamiltonian exactly, completing the analogy between the sublattice pseudospin travelling in the monolayer graphene with ripples and strain fields, and the real spin $1/2$ fermion travelling in a curved space. The effective metric as seen by the sublattice pseudospin is different from the real space metric as defined by the two-dimensional manifold of the graphene monolayer. All relevant terms of the effective gauge field from the microscopic model is calculated exactly for a unimodular effective metric.

cond-mat.mes-hall↗