This paper considers on-line and semi-on-line scheduling problems on m parallel machines with objective to maximize the minimum load. For on-line version, we prove that algorithm Random is an optimal randomized algori...
详细信息
This paper considers on-line and semi-on-line scheduling problems on m parallel machines with objective to maximize the minimum load. For on-line version, we prove that algorithm Random is an optimal randomized algorithm on two machines, and derive a new randomized upper bound for general m machines which significantly improves the known upper bound. For semi-on-line version with non-increasing job processing times, we show that LS algorithm is an optimal deterministic algorithm for two and three machine cases. We further present an optimal randomized algorithm RLS for two machine case.
This paper considers randomized discrete-time consensus systems that preserve the average "on average". As a main result, we provide an upper bound on the mean square deviation of the consensus value from th...
详细信息
This paper considers randomized discrete-time consensus systems that preserve the average "on average". As a main result, we provide an upper bound on the mean square deviation of the consensus value from the initial average. Then, we apply our result to systems in which few or weakly correlated interactions take place: these assumptions cover several algorithms proposed in the literature. For such systems we show that, when the network size grows, the deviation tends to zero, and that the speed of this decay is not slower than the inverse of the size. Our results are based on a new approach, which is unrelated to the convergence properties of the system. (C) 2013 Elsevier Ltd. All rights reserved.
We review the basic outline of the highly successful diffusion Monte Carlo technique commonly used in contexts ranging from electronic structure calculations to rare event simulation and data assimilation, and propose...
详细信息
We review the basic outline of the highly successful diffusion Monte Carlo technique commonly used in contexts ranging from electronic structure calculations to rare event simulation and data assimilation, and propose a new class of randomized iterative algorithms based on similar principles to address a variety of common tasks in numerical linear algebra. From the point of view of numerical linear algebra, the main novelty of the fast randomized iteration schemes described in this article is that they have dramatically reduced operations and storage cost per iteration (as low as constant under appropriate conditions) and are rather versatile: we will show how they apply to the solution of linear systems, eigenvalue problems, and matrix exponentiation, in dimensions far beyond the present limits of numerical linear algebra. While traditional iterative methods in numerical linear algebra were created in part to deal with instances where a matrix (of size O(n(2))) is too big to store, the algorithms that we propose are effective even in instances where the solution vector itself (of size O(n)) may be too big to store or manipulate. In fact, our work is motivated by recent diffusion Monte Carlo based quantum Monte Carlo schemes that have been applied to matrices as large as 10(108) x 10(108). We provide basic convergence results, discuss the dependence of these results on the dimension of the system, and demonstrate dramatic cost savings on a range of test problems.
The problem of document replacement in web caches has received much attention in recent research, and it has been shown that the eviction rule "replace the least recently used document" performs poorly in we...
详细信息
The problem of document replacement in web caches has received much attention in recent research, and it has been shown that the eviction rule "replace the least recently used document" performs poorly in web caches. Instead, it has been shown that using a combination of several criteria, such as the recentness and frequency of use, the size, and the cost of fetching a document, leads to a sizable improvement in hit rate and latency reduction. However, in order to implement these novel schemes, one needs to maintain complicated data structures. We propose randomized algorithms for, approximating any existing web-cache replacement scheme and thereby avoid the need for any data structures. At document-replacement times, the randomized algorithm samples N documents from the cache and replaces the least useful document from the sample, where usefulness is determined according to the criteria mentioned above. The next M < N least useful documents are retained for the succeeding iteration. When the next replacement is to be performed, the algorithm obtains N - M new samples from the cache and replaces the least useful document from the N - M new samples and the M previously retained. Using theory and simulations, we analyze the algorithm and find that it matches the performance of existing document replacement schemes for values of N and M as low as 8 and 2 respectively. Interestingly, we find. that retaining. a small number of samples from one iteration to the next leads to an exponential improvement in performance as compared to retaining no samples at all.
We propose a novel randomized linear programming algorithm for approximating the optimal policy of the discounted-reward and average-reward Markov decision problems. By leveraging the value-policy duality, the algorit...
详细信息
We propose a novel randomized linear programming algorithm for approximating the optimal policy of the discounted-reward and average-reward Markov decision problems. By leveraging the value-policy duality, the algorithm adaptively samples state-action-state transitions and makes exponentiated primal-dual updates. We show that it finds an f-optimal policy using nearly linear runtime in the worst case for a fixed value of the discount factor. When the Markov decision process is ergodic and specified in some special data formats, for fixed values of certain ergodicity parameters, the algorithm finds an c-optimal policy using sample size and time linear in the total number of state-action pairs, which is sublinear in the input size. These results provide a new venue and complexity benchmarks for solving stochastic dynamic programs.
We consider the gathering problem of multiple (mobile) agents in anonymous unidirectional ring networks under the constraint that each agent knows neither the number of nodes nor the number of agents. For this problem...
详细信息
We consider the gathering problem of multiple (mobile) agents in anonymous unidirectional ring networks under the constraint that each agent knows neither the number of nodes nor the number of agents. For this problem, we fully characterize the relation between probabilistic solvability and termination detection. First, we prove for any (small) constant p (0 < p <= 1) that no randomized algorithm exists that solves, with probability p, the gathering problem with (termination) detection. For this reason, we consider the relaxed gathering problem, called the gathering problem without detection, which does not require termination detection. We propose a randomized algorithm that solves, with any given constant probability p (0 < p < 1), the gathering problem without detection. Finally, we prove that no randomized algorithm exists that solves, with probability 1, the gathering problem without detection.
For a fixed integer the hypergraph k-cut problem asks for a smallest subset of hyperedges whose removal leads to at least k connected components in the remaining hypergraph. While graph k-cut is solvable efficiently (...
详细信息
For a fixed integer the hypergraph k-cut problem asks for a smallest subset of hyperedges whose removal leads to at least k connected components in the remaining hypergraph. While graph k-cut is solvable efficiently (Goldschmidt and Hochbaum in Math. Oper. Res. 19(1):24-37, 1994), the complexity of hypergraph k-cut has been open. In this work, we present a randomized polynomial time algorithm to solve the hypergraph k-cut problem. Our algorithmic technique extends to solve the more general hedge k-cut problem when the subgraph induced by every hedge has a constant number of connected components. Our algorithm is based on random contractions akin to Karger's min cut algorithm. Our main technical contribution is a non-uniform distribution over the hedges (hyperedges) so that random contraction of hedges (hyperedges) chosen from the distribution succeeds in returning an optimum solution with large probability. In addition, we present an alternative contraction based randomized polynomial time approximation scheme for hedge k-cut in arbitrary hedgegraphs (i.e., hedgegraphs whose hedges could have a large number of connected components). Our algorithm and analysis also lead to bounds on the number of optimal solutions to the respective problems.
Decomposition of a matrix into low-rank matrices is a powerful tool for scientific computing and data analysis. The purpose is to obtain a low-rank matrix by decomposition of the original matrix into a product of smal...
详细信息
Decomposition of a matrix into low-rank matrices is a powerful tool for scientific computing and data analysis. The purpose is to obtain a low-rank matrix by decomposition of the original matrix into a product of smaller and lower-rank matrices or by randomly projecting the matrix down to a lower-dimensional space. Such decomposition requires less storage and computational burden. The focus of this paper is on randomized methods which try as much as possible to preserve the original matrix properties by applying the subspace sampling. In many applications, randomized algorithms in terms of accuracy, stability and speed are much better than the classical decomposition algorithms. In this study, we propose a sparse orthogonal transformation matrix to reduce the dimension of the data. The results show that compared with the most accurate methods, the transformation speed is much faster and can save a lot of memory in the case of huge matrices.
We present randomized algorithms based on block Krylov subspace methods for estimating the trace and log-determinant of Hermitian positive semi-definite matrices. Using the properties of Chebyshev polynomials and Gaus...
详细信息
We present randomized algorithms based on block Krylov subspace methods for estimating the trace and log-determinant of Hermitian positive semi-definite matrices. Using the properties of Chebyshev polynomials and Gaussian random matrix, we provide the error analysis of the proposed estimators and obtain the expectation and concentration error bounds. These bounds improve the corresponding ones given in the literature. Numerical experiments are presented to illustrate the performance of the algorithms and to test the error bounds.
One popular way to compute the CANDECOMP/PARAFAC (CP) decomposition of a tensor is to transform the problem into a sequence of overdetermined least squares subproblems with Khatri-Rao product (KRP) structure involving...
详细信息
One popular way to compute the CANDECOMP/PARAFAC (CP) decomposition of a tensor is to transform the problem into a sequence of overdetermined least squares subproblems with Khatri-Rao product (KRP) structure involving factor matrices. In this work, based on choosing the factor matrix randomly, we propose a mini-batch stochastic gradient descent method with importance sampling for those special least squares subproblems. Two different sampling strategies are provided. They can avoid forming the full KRP explicitly and computing the corresponding probabilities directly. The adaptive step size version of the method is also given. For the proposed method, we present its theoretical properties and comprehensive numerical performance. The results on synthetic and real data show that our method is effective and efficient, and for unevenly distributed data, it performs better than the corresponding one in the literature.
暂无评论