arXiv subjects

Xiang Wang

Publications and source records attributed to Xiang Wang.

437 records · Page 25Linked to original sources

arXiv2015

Well Control Optimization using Derivative-Free Algorithms and a Multiscale Approach

In this paper, we use numerical optimization algorithms and a multiscale approach in order to find an optimal well management strategy over the life of the reservoir. The large number of well rates for each control step make the optimization problem more difficult and at a high risk of achieving a suboptimal solution. Moreover, the optimal number of adjustments is not known a priori. Adjusting well controls too frequently will increase unnecessary well management and operation cost, and an excessively low number of control adjustments may not be enough to obtain a good yield. We investigate three derivative-free optimization algorithms, chosen for their robust and parallel nature, to determine optimal well control strategies. The algorithms chosen include generalized pattern search (GPS), particle swarm optimization (PSO) and covariance matrix adaptation evolution strategy (CMA-ES). These three algorithms encompass the breadth of available black-box optimization strategies: deterministic local search, stochastic global search and stochastic local search. In addition, we hybridize the three derivative-free algorithms with a multiscale regularization approach. Starting with a reasonably small number of control steps, the control intervals are subsequently refined during the optimization. Results for experiments studied indicate that CMA-ES performs best among the three algorithms in solving both small and large scale problems. When hybridized with a multiscale regularization approach, the ability to find the optimal solution is further enhanced, with the performance of GPS improving the most. Topics affecting the performance of the multiscale approach are discussed in this paper, including the effect of control frequency on the well control problem. The parameter settings for GPS, PSO, and CMA-ES, within the multiscale approach are considered.

math.OC↗

arXiv2015

Effects of systematic phase errors on optimized quantum random-walk search algorithm

This paper researches how the systematic errors in phase inversions affect the success rate and the number of iterations in optimized quantum random-walk search algorithm. Through geometric description of this algorithm, the model of the algorithm with phase errors is established and the relationship between the success rate of the algorithm, the database size, the number of iterations and the phase error is depicted. For a given sized database, we give both the maximum success rate of the algorithm and the required number of iterations when the algorithm is in the presence of phase errors. Through analysis and numerical simulations, it shows that optimized quantum random-walk search algorithm is more robust than Grover's algorithm.

quant-ph↗

arXiv2013

Detecting the rapidly expanding outer shell of the Crab Nebula: where to look

We present a range of steady-state photoionization simulations, corresponding to different assumed shell geometries and compositions, of the unseen postulated rapidly expanding outer shell to the Crab Nebula. The properties of the shell are constrained by the mass that must lie within it, and by limits to the intensities of hydrogen recombination lines. In all cases the photoionization models predict very strong emission from high ionization lines that will not be emitted by the Crab's filaments, alleviating problems with detecting these lines in the presence of light scattered from brighter parts of the Crab. The NIR [Ne VI] $λ$7.652 $μ$m line is a particularly good case; it should be dramatically brighter than the optical lines commonly used in searches. The C IV $\lambda1549Å$ doublet is predicted to be the strongest absorption line from the shell, which is in agreement with HST observations. We show that the cooling timescale for the outer shell is much longer than the age of the Crab, due to the low density. This means that the temperature of the shell will actually "remember" its initial conditions. However, the recombination time is much shorter than the age of the Crab, so the predicted level of ionization should approximate the real ionization. In any case, it is clear that IR observations present the best opportunity to detect the outer shell and so guide future models that will constrain early events in the original explosion.

astro-ph.GA↗

arXiv2012

On Constrained Spectral Clustering and Its Applications

Constrained clustering has been well-studied for algorithms such as $K$-means and hierarchical clustering. However, how to satisfy many constraints in these algorithmic settings has been shown to be intractable. One alternative to encode many constraints is to use spectral clustering, which remains a developing area. In this paper, we propose a flexible framework for constrained spectral clustering. In contrast to some previous efforts that implicitly encode Must-Link and Cannot-Link constraints by modifying the graph Laplacian or constraining the underlying eigenspace, we present a more natural and principled formulation, which explicitly encodes the constraints as part of a constrained optimization problem. Our method offers several practical advantages: it can encode the degree of belief in Must-Link and Cannot-Link constraints; it guarantees to lower-bound how well the given constraints are satisfied using a user-specified threshold; it can be solved deterministically in polynomial time through generalized eigendecomposition. Furthermore, by inheriting the objective function from spectral clustering and encoding the constraints explicitly, much of the existing analysis of unconstrained spectral clustering techniques remains valid for our formulation. We validate the effectiveness of our approach by empirical results on both artificial and real datasets. We also demonstrate an innovative use of encoding large number of constraints: transfer learning via constraints.

cs.LG↗

arXiv2012

A Reconstruction Error Formulation for Semi-Supervised Multi-task and Multi-view Learning

A significant challenge to make learning techniques more suitable for general purpose use is to move beyond i) complete supervision, ii) low dimensional data, iii) a single task and single view per instance. Solving these challenges allows working with "Big Data" problems that are typically high dimensional with multiple (but possibly incomplete) labelings and views. While other work has addressed each of these problems separately, in this paper we show how to address them together, namely semi-supervised dimension reduction for multi-task and multi-view learning (SSDR-MML), which performs optimization for dimension reduction and label inference in semi-supervised setting. The proposed framework is designed to handle both multi-task and multi-view learning settings, and can be easily adapted to many useful applications. Information obtained from all tasks and views is combined via reconstruction errors in a linear fashion that can be efficiently solved using an alternating optimization scheme. Our formulation has a number of advantages. We explicitly model the information combining mechanism as a data structure (a weight/nearest-neighbor matrix) which allows investigating fundamental questions in multi-task and multi-view learning. We address one such question by presenting a general measure to quantify the success of simultaneous learning of multiple tasks or from multiple views. We show that our SSDR-MML approach can outperform many state-of-the-art baseline methods and demonstrate the effectiveness of connecting dimension reduction and learning.

cs.LG↗