arXiv subjects

Shuang Gao

Publications and source records attributed to Shuang Gao.

At least 37 records · Page 2Linked to original sources

arXiv2022

Hardware-algorithm collaborative computing with photonic spiking neuron chip based on integrated Fabry-Pérot laser with saturable absorber

Photonic neuromorphic computing has emerged as a promising avenue toward building a low-latency and energy-efficient non-von-Neuman computing system. Photonic spiking neural network (PSNN) exploits brain-like spatiotemporal processing to realize high-performance neuromorphic computing. However, the nonlinear computation of PSNN remains a significant challenging. Here, we proposed and fabricated a photonic spiking neuron chip based on an integrated Fabry-Pérot laser with a saturable absorber (FP-SA) for the first time. The nonlinear neuron-like dynamics including temporal integration, threshold and spike generation, refractory period, and cascadability were experimentally demonstrated, which offers an indispensable fundamental building block to construct the PSNN hardware. Furthermore, we proposed time-multiplexed spike encoding to realize functional PSNN far beyond the hardware integration scale limit. PSNNs with single/cascaded photonic spiking neurons were experimentally demonstrated to realize hardware-algorithm collaborative computing, showing capability in performing classification tasks with supervised learning algorithm, which paves the way for multi-layer PSNN for solving complex tasks.

cs.ET↗

arXiv2021

LQG Graphon Mean Field Games: Analysis via Graphon Invariant Subspaces

This paper studies approximate solutions to large-scale linear quadratic stochastic games with homogeneous nodal dynamics parameters and heterogeneous network couplings within the graphon mean field game framework in [2]-[4]. A graphon time-varying dynamical system model is first formulated to study the finite and then limit problems of linear quadratic Gaussian graphon mean field games (LQG-GMFG). The Nash equilibrium of the limit problem is then characterized by two coupled graphon time-varying dynamical systems. Sufficient conditions are established for the existence of a unique solution to the limit LQG-GMFG problem. For the computation of LQG-GMFG solutions two methods are established and employed where one is based on fixed point iterations and the other on a decoupling operator Riccati equation; furthermore, two corresponding sets of solutions are established based on spectral decompositions. Finally, a set of numerical simulations on networks associated with different types of graphons are presented.

eess.SY↗

arXiv2021

Retrieval and Localization with Observation Constraints

Accurate visual re-localization is very critical to many artificial intelligence applications, such as augmented reality, virtual reality, robotics and autonomous driving. To accomplish this task, we propose an integrated visual re-localization method called RLOCS by combining image retrieval, semantic consistency and geometry verification to achieve accurate estimations. The localization pipeline is designed as a coarse-to-fine paradigm. In the retrieval part, we cascade the architecture of ResNet101-GeM-ArcFace and employ DBSCAN followed by spatial verification to obtain a better initial coarse pose. We design a module called observation constraints, which combines geometry information and semantic consistency for filtering outliers. Comprehensive experiments are conducted on open datasets, including retrieval on R-Oxford5k and R-Paris6k, semantic segmentation on Cityscapes, localization on Aachen Day-Night and InLoc. By creatively modifying separate modules in the total pipeline, our method achieves many performance improvements on the challenging localization benchmarks.

cs.CV↗

arXiv2021

Centrality-Weighted Opinion Dynamics: Disagreement and Social Network Partition

This paper proposes a network model of opinion dynamics based on both the social network structure and network centralities. The conceptual novelty in this model is that the opinion of each individual is weighted by the associated network centrality in characterizing the opinion spread on social networks. Following a degree-centrality-weighted opinion dynamics model, we provide an algorithm to partition nodes of any graph into two and multiple clusters based on opinion disagreements. Furthermore, the partition algorithm is applied to real-world social networks including the Zachary karate club network [1] and the southern woman network [2] and these application examples indirectly verify the effectiveness of the degree-centrality-weighted opinion dynamics model. Finally, properties of general centrality-weighted opinion dynamics model are established.

cs.SI↗

arXiv2021

Subspace Decomposition for Graphon LQR: Applications to VLSNs of Harmonic Oscillators

Graphon control has been proposed and developed in [1]-[3] to approximately solve control problems for very large-scale networks (VLSNs) of linear dynamical systems based on graphon limits. This paper provides a solution method based on invariant subspace decompositions for a class of graphon linear quadratic regulation (LQR) problems where the local dynamics share homogeneous parameters but the graphon couplings may be heterogeneous among the coupled agents. Graphon couplings in this paper appear in states, controls and costs, and these couplings may be represented by different graphons. By exploring a common invariant subspace of the couplings, the original problem is decomposed into a network coupled LQR problem of finite dimension and a decoupled infinite dimensional LQR problem. A centralized optimal solution, and a nodal collaborative optimal control solution where each agent computes its part of the optimal solution locally, are established. The application of these solutions to finite network LQR problems may be via (i) the graphon control methodology [3], or (ii) the representation of finite LQR problems as special cases of graphon LQR problems. The complexity of these solutions involves solving one nd X nd dimensional Riccati equation and one n X n Riccati equation, where n is the dimension of each nodal agent state and d is the dimension of the nontrivial common invariant subspace of the coupling operators, whereas a direct approach involves solving an nN X nN dimensional Riccati equation, where N is the size of the network. For situations where the graphon couplings do not admit exact low-rank representations, approximate control is developed based on low-rank approximations. Finally, an application to the regulation of harmonic oscillators coupled over large networks with uncertainties is demonstrated.

eess.SY↗

arXiv2021

Optimal control of network-coupled subsystems: Spectral decomposition and low-dimensional solutions

In this paper, we investigate optimal control of network-coupled subsystems where the dynamics and the cost couplings depend on an underlying undirected weighted graph. The graph coupling matrix in the dynamics may be the adjacency matrix, the Laplacian matrix, or any other symmetric matrix corresponding to the underlying graph. The cost couplings can be any polynomial function of the underlying coupling matrix. We use the spectral decomposition of the graph coupling matrix to decompose the overall system into (L+1) systems with decoupled dynamics and cost, where L is the rank of the coupling matrix. Furthermore, the optimal control input at each subsystem can be computed by solving (Ldist + 1) decoupled Riccati equations where Ldist (Ldist \leq L) is the number of distinct non-zero eigenvalues of the coupling matrix. A salient feature of the result is that the solution complexity depends on the number of distinct eigenvalues of the coupling matrix rather than the size of the network. Therefore, the proposed solution framework provides a scalable method for synthesizing and implementing optimal control laws for large-scale network-coupled subsystems.

eess.SY↗

arXiv2020

Self-semi-supervised Learning to Learn from NoisyLabeled Data

The remarkable success of today's deep neural networks highly depends on a massive number of correctly labeled data. However, it is rather costly to obtain high-quality human-labeled data, leading to the active research area of training models robust to noisy labels. To achieve this goal, on the one hand, many papers have been dedicated to differentiating noisy labels from clean ones to increase the generalization of DNN. On the other hand, the increasingly prevalent methods of self-semi-supervised learning have been proven to benefit the tasks when labels are incomplete. By 'semi' we regard the wrongly labeled data detected as un-labeled data; by 'self' we choose a self-supervised technique to conduct semi-supervised learning. In this project, we designed methods to more accurately differentiate clean and noisy labels and borrowed the wisdom of self-semi-supervised learning to train noisy labeled data.

cs.LG↗

arXiv2020

Linear Quadratic Graphon Field Games

Linear quadratic graphon field games (LQ-GFGs) are defined to be LQ games which involve a large number of agents that are weakly coupled via a weighted undirected graph on which each node represents an agent. The links of the graph correspond to couplings between the agents' dynamics, as well as between the individual cost functions, which each agent attempts to minimize. We formulate limit LQ-GFG problems based on the assumption that these graphs lie in a sequence which converges to a limit graphon. First, under a finite-rank assumption on the limit graphon, the existence and uniqueness of solutions to the formulated limit LQ-GFG problem is established. Second, based upon the solutions to the limit LQ-GFG problem, epsilon-Nash equilibria are constructed for the corresponding game problems with a very large but finite number of players. This result is then generalized to the case with random initial conditions. It is to be noted that LQ-GFG problems are distinct from the class of graphon mean field game (GMFG) problems where a population is hypothesized to be associated with each node of the graph [Caines and Huang CDC 2018, 2019].

eess.SY↗

arXiv2020

Visual Localization Using Semantic Segmentation and Depth Prediction

In this paper, we propose a monocular visual localization pipeline leveraging semantic and depth cues. We apply semantic consistency evaluation to rank the image retrieval results and a practical clustering technique to reject estimation outliers. In addition, we demonstrate a substantial performance boost achieved with a combination of multiple feature extractors. Furthermore, by using depth prediction with a deep neural network, we show that a significant amount of falsely matched keypoints are identified and eliminated. The proposed pipeline outperforms most of the existing approaches at the Long-Term Visual Localization benchmark 2020.

cs.CV↗

arXiv2019

Graphon Control of Large-scale Networks of Linear Systems

To achieve control objectives for extremely large-scale complex networks using standard methods is essentially intractable. In this work a theory of the approximate control of complex network systems is proposed and developed by the use of graphon theory and the theory of infinite dimensional systems. First, graphon dynamical system models are formulated in an appropriate infinite dimensional space in order to represent arbitrary-size networks of linear dynamical systems, and to define the convergence of sequences of network systems with limits in the space. Exact controllability and approximate controllability of graphon dynamical systems are then investigated. Second, the minimum energy state-to-state control problem and the linear quadratic regulator problem for systems on complex networks are considered. The control problem for graphon limit systems is solved in each case and approximations are defined which yield control laws for the original control problems. Furthermore, convergence properties of the approximation schemes are established. A systematic control design methodology is developed within this framework. Finally, numerical examples of networks with randomly sampled weightings are presented to illustrate the effectiveness of the graphon control methodology.

math.OC↗

arXiv2019

Optimal and Approximate Solutions to Linear Quadratic Regulation of a Class of Graphon Dynamical Systems

In this paper we study the linear quadratic regulation (LQR) problem for dynamical systems coupled over large-scale networks and obtain locally computable low-complexity solutions. The underlying large or even infinite networks are represented by graphons and the couplings appear in both the dynamics and the quadratic cost. The optimal solution is obtained first for graphon dynamical systems for the special case where the graphons are exactly characterized by finite spectral summands. The complexity of generating these control solutions involves solving d+1 scalar Riccati equations where d is the number of non-zero eigenvalues in the spectral representation. Based on this, we provide a suboptimal low-complexity solution for problems with general graphon couplings via spectral approximations and analyze the performance under the approximate control. Finally, a numerical example is given to illustrate the explicit solution and demonstrate the simplicity of the solution.

math.OC↗

arXiv2019

Spectral Representations of Graphons in Very Large Network Systems Control

Graphon-based control has recently been proposed and developed to solve control problems for dynamical systems on networks which are very large or growing without bound (see Gao and Caines, CDC 2017, CDC 2018). In this paper, spectral representations, eigenfunctions and approximations of graphons, and their applications to graphon-based control are studied. First, spectral properties of graphons are presented and then approximations based on Fourier approximated eigenfunctions are analyzed. Within this framework, two classes of graphons with simple spectral representations are given. Applications to graphon-based control analysis are next presented; in particular, the controllability of systems distributed over very large networks is expressed in terms of the properties of the corresponding graphon dynamical systems. Moreover, spectral analysis based upon real-world network data is presented, which demonstrates that low-dimensional spectral approximations of networks are possible. Finally, an initial, exploratory investigation of the utility of the spectral analysis methodology in graphon systems control to study the control of epidemic spread is presented.

eess.SY↗

arXiv2019

VACL: Variance-Aware Cross-Layer Regularization for Pruning Deep Residual Networks

Improving weight sparsity is a common strategy for producing light-weight deep neural networks. However, pruning models with residual learning is more challenging. In this paper, we introduce Variance-Aware Cross-Layer (VACL), a novel approach to address this problem. VACL consists of two parts, a Cross-Layer grouping and a Variance Aware regularization. In Cross-Layer grouping the $i^{th}$ filters of layers connected by skip-connections are grouped into one regularization group. Then, the Variance-Aware regularization term takes into account both the first and second-order statistics of the connected layers to constrain the variance within a group. Our approach can effectively improve the structural sparsity of residual models. For CIFAR10, the proposed method reduces a ResNet model by up to 79.5% with no accuracy drop and reduces a ResNeXt model by up to 82% with less than 1% accuracy drop. For ImageNet, it yields a pruned ratio of up to 63.3% with less than 1% top-5 accuracy drop. Our experimental results show that the proposed approach significantly outperforms other state-of-the-art methods in terms of overall model size and accuracy.

cs.CV↗

arXiv2018

The Ultraviolet Extinction in the $GALEX$ Bands

Interstellar extinction in ultraviolet is the most severe in comparison with optical and infrared wavebands and a precise determination plays an important role in correctly recovering the ultraviolet brightness and colors of objects. By finding the observed bluest colors at given effective temperature and metallicity range of dwarf stars, stellar intrinsic colors, $C^0_{\rm B,V}$, $C^0_{\rm NUV,B}$, $C^0_{\rm FUV,B}$ and $C^0_{\rm FUV,NUV}$, are derived according to the stellar parameters from the LAMOST spectroscopic survey and photometric results from the $GALEX$ and APASS surveys. With the derived intrinsic colors, the ultraviolet color excesses are calculated for about 25,000 A- and F-type dwarf stars. Analysis of the color excess ratios yields the extinction law related to the $GALEX$ UV bands: $E_{\rm NUV,B}$/$E_{\rm B,V} = 3.77$, $E_{\rm FUV,B}$/$E_{\rm B,V} = 3.39$, $E_{\rm FUV,NUV}$/$E_{\rm B,V} = -0.38$. The results agree very well with previous works in the $NUV$ band and in general with the extinction curve derived by Fitzpatrick (1999) for $R_{\rm V}=3.35$.

astro-ph.SR↗

arXiv2018

Improving the Closed-Loop Tracking Performance Using the First-Order Hold Sensing Technique with Experiments

This paper proposes a new perspective in the enhancement of the closed-loop tracking performance by using the first-order hold (FOH) sensing technique. Firstly, the literature review and fundamentals of the FOH are outlined. Secondly, the performance of the most commonly used zero-order hold (ZOH) and that of the FOH are compared. Lastly, the detailed implementation of the FOH on a pendulum tracking setup is presented to verify the superiority of the FOH over the ZOH in terms of the steady state tracking error. The results of the simulation and the experiment are in agreement.

eess.SY↗

arXiv2017

The binarity of Galactic dwarf stars along with effective temperature and metallicity

The fraction of binary stars (fb) is one of most valuable tool to probe the star formation and evolution of multiple systems in the Galaxy. We focus on the relationship between fb and stellar metallicity ([Fe/H]) by employing the differential radial velocity (DRV) method and the large sample observed by the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST). Main-sequence stars from A- to K-types in the third data release (DR3) of LAMOST are selected to estimate fb. Contributions to a profile of DRV from radial velocity (RV) error of single stars (σ) and orbital motion of binary stars are evaluated from the profile of DRV. Finally, we employ 365,911 stars with randomly repeating spectral observations to present a detailed analysis of fb and σin the two-dimensional (2D) space of Teff and [Fe/H]. The A-type stars are more likely to be companions in binary star systems than other stars. Furthermore, the reverse correlation between fb and [Fe/H] can be shown statistically, which suggests that fb is a joint function of Teff and [Fe/H]. At the same time, σof the sample for different Teff and [Fe/H] are fitted. Metal-rich cold stars in our sample have the best RV measurement.

astro-ph.SR↗

arXiv2016

Revision of Stellar Intrinsic Colors in the Infrared by the Spectroscopic Surveys

Intrinsic colors of normal stars are derived in the popularly used infrared bands involving the 2MASS/JHKs, WISE, Spitzer/IRAC and AKARI/S9W filters. Based on three spectroscopic surveys -- LAMOST, RAVE and APOGEE, stars are classified into groups of giants and dwarfs, as well as metal-normal and metal-poor stars. An empirical analytical relation of the intrinsic color is obtained with stellar effective temperature (Teff) for each group of stars after the zero-reddening stars are selected from the blue edge in the $J-λ$ versus (Teff) diagram. It is found that metallicity has little effect on the infrared colors. In the near-infrared bands, our results agree with previous work. In addition, the color indexes H-W2 and Ks-W1 that are taken as constant to calculate interstellar extinction are discussed. The intrinsic color of M-type stars are derived separately due to lack of accurate measurement of their effective temperature.

astro-ph.SR↗

arXiv2016

Sharp Chandra View of ROSAT All-Sky Survey Bright Sources: I. Improvement of Positional Accuracy

The ROSAT All-Sky Survey (RASS) represents one of the most complete and sensitive soft X-ray all-sky surveys to date. However, the deficient positional accuracy of the RASS Bright Source Catalog (BSC) and subsequent lack of firm optical identifications affect the multi-wavelength studies of X-ray sources. The widely used positional errors $σ_{pos}$ based on the Tycho Stars Catalog (Tycho-1) have previously been applied for identifying objects in the optical band. The considerably sharper Chandra view covers a fraction of RASS sources, whose $σ_{pos}$ could be improved by utilizing the sub-arcsec positional accuracy of Chandra observations. We cross-match X-ray objects between the BSC and \emph{Chandra} sources extracted from the Advanced CCD Imaging Spectrometer (ACIS) archival observations. A combined counterparts list (BSCxACIS) with \emph{Chandra} spatial positions weighted by the X-ray flux of multi-counterparts is employed to evaluate and improve the former identifications of BSC with the other surveys. Based on these identification evaluations, we suggest that the point-likeness of BSC sources and INS (isolated neutron stars) candidates should be carefully reconsidered.

astro-ph.HE↗