Recently, the decentralized optimization problem is attracting growing attention. Most existing methods are deterministic with high per-iteration cost and have a convergence rate quadratically depending on condition number. Besides, dense communication necessary to ensure even if dataset sparse. In this paper, we generalize monotone operator root finding problem, propose stochastic algorithm named DSBA that (i) converges geometrically linearly number, (ii) can be implemented using sparse...

Multipath is the main systematic error of Global Navigation Satellite System (GNSS) short baseline positioning. cannot be eliminated by double-differenced technique and difficult to parameterize, which severely restrict high-precision GNSS positioning application. Based on spatiotemporal repeatability multipath, sidereal filtering in coordinate-domain (SF-CD), observation-domain (SF-OD), multipath hemispherical map (MHM) can used mitigate real-time. However, model with large matrix for...

The BeiDou Navigation Satellite System (BDS) features a heterogeneous constellation so that it is difficult to mitigate the multipath in coordinate-domain. Therefore, mitigating observation-domain becomes more important. Sidereal filtering commonly used for mitigation, which needs calculate orbit repeat time of each satellite. However, poses computational challenge and damages integrity at end model. this paper proposes single-difference model based on hemispherical map (SD-MHM) BDS-2/BDS-3...

By restricting the iterate on a nonlinear manifold, recently proposed Riemannian optimization methods prove to be both efficient and effective in low rank tensor completion problems. However, existing fail exploit easily accessible side information, due their format mismatch. Consequently, there is still room for improvement. To fill gap, this paper, novel model tightly integrate original information by overcoming inconsistency. For model, an conjugate gradient descent solver devised based...

In real-world recommendation tasks, feedback data are usually sparse. Therefore, a recommender’s performance is often determined by how much information that it can extract from textual contents. However, current methods do not make full use of the semantic information. They encode contents either “bag-of-words” technique or Recurrent Neural Network (RNN). The former neglects order words while latter ignores fact contain multiple topics. Besides, there exists dilemma in designing...

The number of fragments and the variety primitive cultural relics unearthed in archaeology, especially mixed several dynasties Qinglong town, Shanghai, pose a great challenge to manual splicing. traditional comparison method is easy cause second damage relics. In this paper, edge feature extracted based on removing noise point cloud, bilateral filtering cloud denoising algorithm salient features proposed. By changing step size field view, Improved Artificial Fish Swarm Algorithm used get...

With the recent proliferation of recommendation system, there have been a lot interests in session-based prediction methods, particularly those based on Recurrent Neural Network (RNN) and their variants. However, existing methods either ignore dwell time that plays an important role measuring user's engagement content, or fail to process very short noisy sessions. In this paper, we propose joint predictor, JUMP, for both user click settings. To map its input into feature vector, JUMP adopts...

Due to the existence of environmental or human factors, and because instrument itself, there are many uncertainties in point clouds, which directly affect data quality accuracy subsequent processing, such as cloud segmentation, 3D modeling, etc. In this paper, address problem, stochastic information coordinates is taken into account, on basis scanner observation principle within Gauss–Helmert model, a novel general point-based self-calibration method developed for terrestrial laser scanners,...

Since a vehicle-borne light detection and ranging (LiDAR) measurement system is affected by the signal shielding rolling vibration of vehicle as it moves, commonly available trajectory data are usually low-quality with noise.Although position attitude processed joint Kalman filtering, there still fluctuations in local areas, which require processing to smooth acquired data.In this paper, model motion proposed analyze trend over time recording position, velocity, information real time.Next,...

Fundamental systematic errors in point cloud data are inevitable due to a variety of factors, ranging from the external environment during scanning or observation by terrestrial laser scanner (TLS), assembly instrument. For low-cost scanners, error terms may be further accentuated and include, addition errors, random even serious that directly affect coordinates each cloud, which related quality subsequent processing. To address above issues, we attempted propose robust target-based...

Association rules mining has been under great attention and considered as one of momentous area in data mining. Classical association approaches make implicit assumption that items' importance is the same set a single support for all items. This paper presents an efficient approach users' interest weighted frequent patterns from transactional database. Our paradigm to assign appropriate minimum (minsup) weight each item, which reduces number unnecessary patterns. Furthermore, we also extend...

Alternating Direction Method of Multipliers (ADMM) is a popular method for solving large-scale Machine Learning problems. Stochastic ADMM was proposed to reduce the per iteration computational complexity, which more suitable big data Recently, variance reduction techniques have been integrated with stochastic in order get faster convergence rate, such as SAG-ADMM and SVRG-ADMM. However, their rate still suboptimal w.r.t smoothness constant. In this paper, we propose an accelerated algorithm...

By restricting the iterate on a nonlinear manifold, recently proposed Riemannian optimization methods prove to be both efficient and effective in low rank tensor completion problems. However, existing fail exploit easily accessible side information, due their format mismatch. Consequently, there is still room for improvement such methods. To fill gap, this paper, novel model organically integrate original information by overcoming inconsistency. For particular model, an conjugate gradient...

Implementing effective load recovery strategies can minimize the impact on power grid during failure incidents. This paper proposes a vehicle-to-grid based strategy for distribution network outages. A bi-level model is developed to capture interactions between system operator and shared EV fleet. At upper level, energy price designed by taking into account significance of electrical loads. Both safety economic aspects loads are considered. lower fleet decides whether participate in service...

Algorithms with fast convergence, small number of data access, and low per-iteration complexity are particularly favorable in the big era, due to demand for obtaining \emph{highly accurate solutions} problems \emph{a large samples} \emph{ultra-high} dimensional space. Existing algorithms lack at least one these qualities, thus inefficient handling such challenge. In this paper, we propose a method enjoying all merits an accelerated convergence rate $O(\frac{1}{k^2})$. Empirical studies on...

In this paper, we explore a general Aggregated Gradient Langevin Dynamics framework (AGLD) for the Markov Chain Monte Carlo (MCMC) sampling. We investigate nonasymptotic convergence of AGLD with unified analysis different data accessing (e.g. random access, cyclic access and reshuffle) snapshot updating strategies, under convex nonconvex settings respectively. It is first time that bounds I/O friendly strategies such as reshuffle have been established in MCMC literature. The theoretic...

Large-scale Nuclear Norm penalized Least Square problem (NNLS) is frequently encountered in estimation of low rank structures. In this paper we accelerate the solution procedure by combining non-smooth convex optimization with smooth Riemannian method. Our methods comprise two phases. first phase, use Alternating Direction Method Multipliers (ADMM) both to identify fix manifold where an optimum resides and provide initializer for subsequent refinement. second superlinearly convergent...

The cooperation of rebalancing and vehicle-to-grid services by electric vehicle (EV) drivers can help alleviate traffic congestion provide flexible power resources. This paper introduces a game model where fleet EV make both routing discharging decisions to maximize their own utility. utility is affected various factors, including travel time, Moreover, are also constrained the road capacity charging station capacity. An aggregative established describe non-cooperative behavior drivers, each...