arXiv subjects

Bin Hu

Publications and source records attributed to Bin Hu.

At least 271 records · Page 15Linked to original sources

arXiv2019

Supervised feature selection with orthogonal regression and feature weighting

Effective features can improve the performance of a model, which can thus help us understand the characteristics and underlying structure of complex data. Previous feature selection methods usually cannot keep more local structure information. To address the defects previously mentioned, we propose a novel supervised orthogonal least square regression model with feature weighting for feature selection. The optimization problem of the objection function can be solved by employing generalized power iteration (GPI) and augmented Lagrangian multiplier (ALM) methods. Experimental results show that the proposed method can more effectively reduce the feature dimensionality and obtain better classification results than traditional feature selection methods. The convergence of our iterative method is proved as well. Consequently, the effectiveness and superiority of the proposed method are verified both theoretically and experimentally.

cs.LG↗

arXiv2019

Identifying Illicit Accounts in Large Scale E-payment Networks -- A Graph Representation Learning Approach

Rapid and massive adoption of mobile/ online payment services has brought new challenges to the service providers as well as regulators in safeguarding the proper uses such services/ systems. In this paper, we leverage recent advances in deep-neural-network-based graph representation learning to detect abnormal/ suspicious financial transactions in real-world e-payment networks. In particular, we propose an end-to-end Graph Convolution Network (GCN)-based algorithm to learn the embeddings of the nodes and edges of a large-scale time-evolving graph. In the context of e-payment transaction graphs, the resultant node and edge embeddings can effectively characterize the user-background as well as the financial transaction patterns of individual account holders. As such, we can use the graph embedding results to drive downstream graph mining tasks such as node-classification to identify illicit accounts within the payment networks. Our algorithm outperforms state-of-the-art schemes including GraphSAGE, Gradient Boosting Decision Tree and Random Forest to deliver considerably higher accuracy (94.62% and 86.98% respectively) in classifying user accounts within 2 practical e-payment transaction datasets. It also achieves outstanding accuracy (97.43%) for another biomedical entity identification task while using only edge-related information.

cs.SI↗

arXiv2019

Constraints on the cosmic distance duality relation with simulated data of gravitational waves from the Einstein Telescope

The cosmic distance duality relation (CDDR) has been test through several astronomical observations in the last years. This relation establishes a simple equation relating the angular diameter ($D_A$) and luminosity ($D_L$) distances at a redshift $z$, $D_LD_A^{-1}(1+z)^{-2}=η=1$. However, only very recently this relation has been observationally tested at high redshifts ($z \approx 3.6$) by using luminosity distances from type Ia supernovae (SNe Ia) and gamma ray bursts (GRBs) plus angular diameter distances from strong gravitational lensing (SGL) observations. The results show that no significant deviation from the CDDR validity has been verified. In this work, we test the potentialities of future luminosity distances from gravitational waves (GWs) sources to impose limit on possible departures of CDDR jointly with current SGL observations. The basic advantage of $D_L$ from GWs is being insensitive to non-conservation of the number of photons. By simulating 600, 900 and 1200 data of GWs using the Einstein Telescope (ET) as reference, we derive limits on $η(z)$ function and obtain that the results will be at least competitive with current limits from the SNe Ia $+$ GRBs $+$ SGLs analyses.

astro-ph.CO↗

arXiv2018

Local Probabilistic Model for Bayesian Classification: a Generalized Local Classification Model

In Bayesian classification, it is important to establish a probabilistic model for each class for likelihood estimation. Most of the previous methods modeled the probability distribution in the whole sample space. However, real-world problems are usually too complex to model in the whole sample space; some fundamental assumptions are required to simplify the global model, for example, the class conditional independence assumption for naive Bayesian classification. In this paper, with the insight that the distribution in a local sample space should be simpler than that in the whole sample space, a local probabilistic model established for a local region is expected much simpler and can relax the fundamental assumptions that may not be true in the whole sample space. Based on these advantages we propose establishing local probabilistic models for Bayesian classification. In addition, a Bayesian classifier adopting a local probabilistic model can even be viewed as a generalized local classification model; by tuning the size of the local region and the corresponding local model assumption, a fitting model can be established for a particular classification problem. The experimental results on several real-world datasets demonstrate the effectiveness of local probabilistic models for Bayesian classification.

cs.LG↗

arXiv2018

Local Distribution in Neighborhood for Classification

The k-nearest-neighbor method performs classification tasks for a query sample based on the information contained in its neighborhood. Previous studies into the k-nearest-neighbor algorithm usually achieved the decision value for a class by combining the support of each sample in the neighborhood. They have generally considered the nearest neighbors separately, and potentially integral neighborhood information important for classification was lost, e.g. the distribution information. This article proposes a novel local learning method that organizes the information in the neighborhood through local distribution. In the proposed method, additional distribution information in the neighborhood is estimated and then organized; the classification decision is made based on maximum posterior probability which is estimated from the local distribution in the neighborhood. Additionally, based on the local distribution, we generate a generalized local classification form that can be effectively applied to various datasets through tuning the parameters. We use both synthetic and real datasets to evaluate the classification performance of the proposed method; the experimental results demonstrate the dimensional scalability, efficiency, effectiveness and robustness of the proposed method compared to some other state-of-the-art classifiers. The results indicate that the proposed method is effective and promising in a broad range of domains.

cs.LG↗

arXiv2018

Deep Levels and Mixed Conductivity in Organometallic Halide Perovskites

Understanding the type, formation energy and capture cross section of defects is one of the challenges in the field of organometallic halide perovskite (OMHP) devices. Currently, such understanding is limited, restricting the power conversion efficiencies of OMHPs solar cells from reaching their Shockley Queisser limit. Here, we report on deep level (DL) defects and their effect on free charge transport properties of single crystalline methylammonium lead bromide perovskite (MAPB). In order to determine DL activation energy and capture cross section we used photo-Hall effect spectroscopy (PHES) with enhanced illumination in both steady-state and dynamic regimes. This method has shown to be convenient due to the direct DL visualization by sub-bandgap photo-excitation of trapped carriers. DLs with activation energies of 1.05 eV, 1.5 eV, and 1.9 eV above valence band were detected. The hole capture cross section was found using photoconductivity relaxation after sub-bandgap photo-excitation. We found the DL defects responsible for non-radiative recombination and its impact on band alignment for the first time. Additionally, the transport properties of MAPB single crystal is measured by Time of Flight (ToF) at several biases. The analysis of ToF measurement further confirms increase of Hall mobility and the enhancement of hole transport produced by sub-bandgap illumination in MAPB devices. Our studies provide a strong evidence on deep levels in OMHPs and opens a richer picture of the role and properties of deep levels in MAPB single crystals as a system model for the first time. The deeper knowledge of the electrical structure of OMHP could open further opportunities in the development of more feasible technology.

cond-mat.mtrl-sci↗

arXiv2018

Co-evolution of supermassive black holes with galaxies from semi-analytic model: stochastic gravitational wave background and black hole clustering

We study the co-evolution of supermassive black holes (SMBHs) with galaxies by means of semi-analytic model (SAM) of galaxy formation based on sub-halo merger trees built from Millennium and Millennium-II simulation. We utilize the simulation results from Guo 2013 and Henriques 2015 to study two aspects of the co-evolution, \emph{i.e.} the stochastic gravitational wave (GW) background generated by SMBH merger and the SMBH/galaxy clustering. The characteristic strain amplitude of GW background predicted by Guo 2013 and Henriques 2015 models are $A_{yr^{-1}}=5.00\times10^{-16}$ and $A_{yr^{-1}}=9.42\times10^{-17}$, respectively. We find the GW amplitude is very sensitive to the galaxy merger rate. The difference in the galaxy merger rate between Guo 2013 and Henriques 2015, results in a factor $5$ deviation in the GW strain amplitude. For clusterings, we calculate the spatially isotropic two point auto- and cross-correlation functions (2PCFs) for both SMBHs and galaxies by using the mock catalogs generated from Guo 2013 model. We find that all 2PCFs have positive dependence on both SMBH and galaxy mass. And there exist a significant time evolution in 2PCFs, namely, the clustering effect is enhanced at lower redshifts. Interestingly, this result is not reported in the active galactic nuclei samples in SDSS. Our analysis also shows that, roughly, SMBHs and galaxies, with galaxy mass $10^2\sim10^3$ larger than SMBH mass, have similar pattern of clustering, which is a reflection of the co-evolution of SMBH and galaxy. Finally, we calculate the first ten multiples of the angular power spectrum of the energy density of GW background. We find the amplitude of angular power spectrum of the first ten multiples is about $10\%$ to $60\%$ of the monopole component in the whole frequency range.

astro-ph.CO↗

arXiv2018

Creation of bipartite steering correlations by a fast damped auxiliary mode

We consider a three-mode system and show how steering correlations can be created between two modes of the system using the fast dissipation of the third mode. These correlations result in a directional form of entanglement, called quantum or EPR steering. We illustrate this on examples of the interactions among damped radiation modes in an optomechanical three-mode system. By assuming that one of the modes undergoes fast dissipation, we show that the coupling of that mode to one or two other modes of the system may result in one- or two-way quantum steering. Explicit analytical results are given for the steering parameters. We find that two modes coupled by the parametric-type interaction and damped with the same rates can be entangled but cannot exhibit quantum steering. When, in addition, one of the modes is coupled to a fast damped mode, steering correlations are created and the modes then exhibit one-way steering. The creation of the steering correlations is interpreted in the context of the variances of the quadrature components of the modes that the steering correlations result from an asymmetry in the variances of the quadrature components of the modes induced by the auxiliary mode. It is found that the fluctuations act directionally that quantum steering may occur only when the variance of the steering mode is larger that the variance of the steered mode. The scheme is shown to be quite robust against the thermal excitation of the modes if the fluctuations of the steering mode are larger than the fluctuations of the steered mode.

quant-ph↗

arXiv2018

Optimal Co-design of Industrial Networked Control Systems with State-dependent Correlated Fading Channels

This paper examines a co-design problem for industrial networked control systems (NCS) whereby physical systems are controlled over wireless fading channels. In particular, the considered wireless channels are also stochastically dependent on the physical states of moving machineries in an industrial working space. In this paper, the moving machineries are modeled as Markov decision processes whereas the characteristics of the correlated fading channels are modeled as a binary random process whose probability measure is a function of both the moving machineries' physical states and the channels' transmission power. Under such state-dependent fading channel models, sufficient conditions which ensure the stochastic safety of the NCS are first derived. Using the derived safety conditions, the co-design problem is then formulated as a constrained joint optimization problem that seeks for optimal control and transmission power policies which simultaneously minimize both the communication and control costs in an infinite time horizon. This paper shows that such optimal co-design policies can be obtained in an efficient manner from the solution of convex programs. Simulation results from industrial NCS models consisting forklift truck and networked DC motor systems are also presented to illustrate and verify both the advantages and efficacy of the proposed co-design solution framework.

math.OC↗

arXiv2018

Dissipativity Theory for Accelerating Stochastic Variance Reduction: A Unified Analysis of SVRG and Katyusha Using Semidefinite Programs

Techniques for reducing the variance of gradient estimates used in stochastic programming algorithms for convex finite-sum problems have received a great deal of attention in recent years. By leveraging dissipativity theory from control, we provide a new perspective on two important variance-reduction algorithms: SVRG and its direct accelerated variant Katyusha. Our perspective provides a physically intuitive understanding of the behavior of SVRG-like methods via a principle of energy conservation. The tools discussed here allow us to automate the convergence analysis of SVRG-like methods by capturing their essential properties in small semidefinite programs amenable to standard analysis and computational techniques. Our approach recovers existing convergence results for SVRG and Katyusha and generalizes the theory to alternative parameter choices. We also discuss how our approach complements the linear coupling technique. Our combination of perspectives leads to a better understanding of accelerated variance-reduced stochastic methods for finite-sum problems.

math.OC↗

arXiv2018

Do current cosmological observations rule out all Covariant Galileons?

We revisit the cosmology of Covariant Galileon gravity in view of the most recent cosmological data sets, including weak lensing. As a higher derivative theory, Covariant Galileon models do not have a $Λ$CDM limit and predict a very different structure formation pattern compared with the standard $Λ$CDM scenario. Previous cosmological analyses suggest that this model is marginally disfavoured, yet can not be completely ruled out. In this work we use a more recent and extended combination of data, and we allow for more freedom in the cosmology, by including a massive neutrino sector with three different mass hierarchies. We use the Planck measurements of Cosmic Microwave Background temperature and polarization; Baryonic Acoustic Oscillations measurements by BOSS DR12; local measurements of $H_0$; the joint light-curve analysis supernovae sample; and, for the first time, weak gravitational lensing from the KiDS collaboration. We find, that in order to provide a reasonable fit, a non-zero neutrino mass is indeed necessary, but we do not report any sizable difference among the three neutrino hierarchies. Finally, the comparison of the Bayesian Evidence to the $Λ$CDM one shows that in all the cases considered, Covariant Galileon models are statistically ruled out by cosmological data.

astro-ph.CO↗

arXiv2018

Concave/convex switchable lens using active phase-change material Ge3Sb2Te6

Normally, the focal length of a conventional lens is fixed. Scientists have made much effort in modulating it into bifocal, which is very important for virtual reality (VR) and argument reality (AR) 3D display. It is even much more difficult for a lens to realize both convex and concave functions with one geometric structure, that is, a concave/convex switchable lens, which can tune 3D real-images and virtual-images in AR and VR, corresponding to long depth of view in 3D display. Based on the tunable refractive indexes of phase-change materials, here we propose a series of concave/convex switchable lenses. When the phase-change material is in different states, one switchable lens is able to perform a negative or positive focal length, or perform negative and positive focal lengths simultaneously. The lenses can be either cylindrical, spherical or other types. For the superior characteristics, these switchable lenses can be employed in various optical fields.

physics.optics↗

arXiv2018

A Robust Accelerated Optimization Algorithm for Strongly Convex Functions

This work proposes an accelerated first-order algorithm we call the Robust Momentum Method for optimizing smooth strongly convex functions. The algorithm has a single scalar parameter that can be tuned to trade off robustness to gradient noise versus worst-case convergence rate. At one extreme, the algorithm is faster than Nesterov's Fast Gradient Method by a constant factor but more fragile to noise. At the other extreme, the algorithm reduces to the Gradient Method and is very robust to noise. The algorithm design technique is inspired by methods from classical control theory and the resulting algorithm has a simple analytical form. Algorithm performance is verified on a series of numerical simulations in both noise-free and relative gradient noise cases.

math.OC↗

arXiv2018

Mapping fast evolution of transient surface photovoltage dynamics using G-Mode Kelvin probe force microscopy

Optoelectronic phenomena in materials such as organic/inorganic hybrid perovskites depend on a complex interplay between light induced carrier generation and fast (electronic) and slower (ionic) processes, all of which are known to be strongly affected by structural inhomogeneities such as interfaces and grain boundaries. Here, we develop a time resolved Kelvin probe force microscopy (KPFM) approach, based on the G-Mode SPM platform, allowing quantification of surface photovoltage (SPV) with microsecond temporal and nanoscale spatial resolution. We demonstrate the approach on methylammonium lead bromide (MAPbBr3) thin films and further highlight the usefulness of unsupervised clustering methods to quickly discern spatial variability in the information rich SPV dataset. Using this technique, we observe concurrent spatial and ultra-fast temporal variations in the SPV generated across the thin film, indicating that structure is likely responsible for the heterogenous behavior.

physics.app-ph↗

arXiv2018

Learning Intrinsic Sparse Structures within Long Short-Term Memory

Model compression is significant for the wide adoption of Recurrent Neural Networks (RNNs) in both user devices possessing limited resources and business clusters requiring quick responses to large-scale service requests. This work aims to learn structurally-sparse Long Short-Term Memory (LSTM) by reducing the sizes of basic structures within LSTM units, including input updates, gates, hidden states, cell states and outputs. Independently reducing the sizes of basic structures can result in inconsistent dimensions among them, and consequently, end up with invalid LSTM units. To overcome the problem, we propose Intrinsic Sparse Structures (ISS) in LSTMs. Removing a component of ISS will simultaneously decrease the sizes of all basic structures by one and thereby always maintain the dimension consistency. By learning ISS within LSTM units, the obtained LSTMs remain regular while having much smaller basic structures. Based on group Lasso regularization, our method achieves 10.59x speedup without losing any perplexity of a language modeling of Penn TreeBank dataset. It is also successfully evaluated through a compact model with only 2.69M weights for machine Question Answering of SQuAD dataset. Our approach is successfully extended to non- LSTM RNNs, like Recurrent Highway Networks (RHNs). Our source code is publicly available at https://github.com/wenwei202/iss-rnns

cs.LG↗

arXiv2018

On $σ$-quasinormal subgroups of finite groups

Let $G$ be a finite group and $σ=\{σ_{i} | i\in I\}$ some partition of the set of all primes $\Bbb{P}$, that is, $σ=\{σ_{i} | i\in I \}$, where $\Bbb{P}=\bigcup_{i\in I} σ_{i}$ and $σ_{i}\cap σ_{j}= \emptyset $ for all $i\ne j$. We say that $G$ is $σ$-primary if $G$ is a $σ_{i}$-group for some $i$. A subgroup $A$ of $G$ is said to be: $σ$-subnormal in $G$ if there is a subgroup chain $A=A_{0} \leq A_{1} \leq \cdots \leq A_{n}=G$ such that either $A_{i-1}\trianglelefteq A_{i}$ or $A_{i}/(A_{i-1})_{A_{i}}$ is $σ$-primary for all $i=1, \ldots, n$, modular in $G$ if the following conditions hold: (i) $\langle X, A \cap Z \rangle=\langle X, A \rangle \cap Z$ for all $X \leq G, Z \leq G$ such that $X \leq Z$, and (ii) $\langle A, Y \cap Z \rangle=\langle A, Y \rangle \cap Z$ for all $Y \leq G, Z \leq G$ such that $A \leq Z$. In this paper, a subgroup $A$ of $G$ is called $σ$-quasinormal in $G$ if $L$ is modular and $σ$-subnormal in $G$. We study $σ$-quasinormal subgroups of $G$. In particular, we prove that if a subgroup $H$ of $G$ is $σ$-quasinormal in $G$, then for every chief factor $H/K$ of $G$ between $H^{G}$ and $H_{G}$ the semidirect product $(H/K)\rtimes (G/C_{G}(H/K))$ is $σ$-primary.

math.GR↗

arXiv2017

CHAM: a fast algorithm of modelling non-linear matter power spectrum in the sCreened HAlo Model

We present a fast numerical screened halo model algorithm (CHAM) for modeling non-linear power spectrum for the alternative models to LCDM. This method has three obvious advantages. First of all, it is not being restricted to a specific dark energy/modified gravity model. In principle, all of the screened scalar-tensor theories can be applied. Second, the least assumptions are made in the calculation. Hence, the physical picture is very easily understandable. Third, it is very predictable and does not rely on the calibration from N-body simulation. As an example, we show the case of Hu-Sawicki f(R) gravity. In this case, the typical CPU time with the current parallel Python script (8 threads) is roughly within $10$ minutes. The resulting spectra are in a good agreement with N-body data within a few percentage accuracy up to k~1 h/Mpc.

astro-ph.CO↗

arXiv2017

Self-triggering in Vehicular Networked Systems with State-dependent Bursty Fading Channels

Vehicular Networked Systems (VNS) are mobile ad hoc networks where vehicles exchange information over wireless communication networks to ensure safe and efficient operation. It is, however, challenging to ensure system safety and efficiency as the wireless channels in VNS are often subject to state-dependent deep fades where the data rate suffers a severe drop and changes as a function of vehicle states. Such couplings between vehicle states and channel states in VNS thereby invalidate the use of separation principle to design event-based control strategies. By adopting a state-dependent fading channel model that was proposed to capture the interaction between vehicle and channel states, this paper presents a novel self-triggered scheme under which the VNS ensures efficient use of communication bandwidth while preserving stochastic stability. The novelty of the proposed scheme lies in its use of the state-dependent fading channel model in the event design that enables an adaptive and effective adjustment on transmission frequency in response to dynamic variations on channel and vehicle states. Under the proposed self-triggered scheme, this paper presents a novel source coding scheme that tracks vehicle's states with performance guarantee in the presence of state-dependent fading channels. The efficacy and advantages of the proposed scheme over other event-based strategies are verified through both simulation and experimental results of a leader-follower example.

eess.SY↗