Tabu Guided Generalized Hill Climbing Algorithms

This paper formulates tabu search strategies that guide generalized hill climbing (GHC) algorithms for addressing NP-hard discrete optimization problems. The resulting framework, termed tabu guided generalized hill climbing (TG²HC) algorithms, uses a tabu release parameter that probabilistically accepts solutions currently on the tabu list. TG²HC algorithms are modeled as a set of stationary Markov chains, where the tabu list is fixed for each outer loop iteration. This framework provides practitioners with guidelines for developing tabu search strategies to use in conjunction with GHC algorithms that preserve some of the algorithms known performance properties. In particular, sufficient conditions are obtained that indicate how to design iterations of problem-specific tabu search strategies, where the stationary distributions associated with each of these iterations converge to the distribution with zero weight on all non-optimal solutions.     

The Gibbs Cloner for Combinatorial Optimization, Counting and Sampling

We present a randomized algorithm, called the cloning algorithm, for approximating the solutions of quite general NP-hard combinatorial optimization problems, counting, rare-event estimation and uniform sampling on complex regions. Similar to the algorithms of Diaconis–Holmes–Ross and Botev–Kroese the cloning algorithm is based on the MCMC (Gibbs) sampler equipped with an importance sampling pdf and, as usual for randomized algorithms, it uses a sequential sampling plan to decompose a “difficult” problem into a sequence of “easy” ones. The cloning algorithm combines the best features of the Diaconis–Holmes–Ross and the Botev–Kroese. In addition to some other enhancements, it has a special mechanism, called the “cloning” device, which makes the cloning algorithm, also called the Gibbs cloner fast and accurate. We believe that it is the fastest and the most accurate randomized algorithm for counting known so far. In addition it is well suited for solving problems associated with the Boltzmann distribution, like estimating the partition functions in an Ising model. We also present a combined version of the cloning and cross-entropy (CE) algorithms. We prove the polynomial complexity of a particular version of the Gibbs cloner for counting. We finally present efficient numerical results with the Gibbs cloner and the combined version, while solving quite general integer and combinatorial optimization problems as well as counting ones, like SAT.     

求解最小Steiner树的蚁群优化算法及其收敛性

最小Steiner树问题是NP难问题,它在通信网络等许多实际问题中有着广泛的应用．蚁群优化算法是最近提出的求解复杂组合优化问题的启发式算法．本文以无线传感器网络中的核心问题之一,路由问题为例,给出了求解最小Steiner树的蚁群优化算法的框架．把算法的迭代过程看作是离散时间的马尔科夫过程,证明了在一定的条件下,该算法所产生的解能以任意接近于1的概率收敛到路由问题的最优解．     

Dual extragradient algorithms extended to equilibrium problems

In this paper we propose two iterative schemes for solving equilibrium problems which are called dual extragradient algorithms. In contrast with the primal extragradient methods in Quoc et al. (Optimization 57(6):749–776, 2008) which require to solve two general strongly convex programs at each iteration, the dual extragradient algorithms proposed in this paper only need to solve, at each iteration, one general strongly convex program, one projection problem and one subgradient calculation. Moreover, we provide the worst case complexity bounds of these algorithms, which have not been done in the primal extragradient methods yet. An application to Nash-Cournot equilibrium models of electricity markets is presented and implemented to examine the performance of the proposed algorithms.     

The use of dynamic programming in genetic algorithms for permutation problems

To deal with computationally hard problems, approximate algorithms are used to provide reasonably good solutions in practical time. Genetic algorithms are an example of the meta-heuristics which were recently introduced and which have been successfully applied to a variety of problems. We propose to use dynamic programming in the process of obtaining new generation solutions in the genetic algorithm, and call it a genetic DP algorithm. To evaluate the effectiveness of this approach, we choose three representative combinatorial optimization problems, the single machine scheduling problem, the optimal linear arrangement problem and the traveling salesman problem, all of which ask to compute optimum permutations of n objects and are known to be NP-hard. Computational results for randomly generated problem instances exhibit encouraging features of genetic DP algorithms.     

Some Randomized Algorithms for Convex Quadratic Programming

We adapt some randomized algorithms of Clarkson [3] for linear programming to the framework of so-called LP-type problems, which was introduced by Sharir and Welzl [10]. This framework is quite general and allows a unified and elegant presentation and analysis. We also show that LP-type problems include minimization of a convex quadratic function subject to convex quadratic constraints as a special case, for which the algorithms can be implemented efficiently, if only linear constraints are present. We show that the expected running times depend only linearly on the number of constraints, and illustrate this by some numerical results. Even though the framework of LP-type problems may appear rather abstract at first, application of the methods considered in this paper to a given problem of that type is easy and efficient. Moreover, our proofs are in fact rather simple, since many technical details of more explicit problem representations are handled in a uniform manner by our approach. In particular, we do not assume boundedness of the feasible set as required in related methods. Accepted 7 May 1997     

Polyhedral studies on the convex recoloring problem

A coloring of the vertices of a graph G is convex if, for each assigned color d, the vertices with color d induce a connected subgraph of G. We address the convex recoloring problem, defined as follows. Given a graph G and a coloring of its vertices, recolor a minimum number of vertices of G, so that the resulting coloring is convex. This problem is known to be NP-hard even when G is a path. We show an integer programming formulation for the weighted version of this problem on arbitrary graphs, and then specialize it for trees. We study the facial structure of the polytope defined as the convex hull of the integer points satisfying the restrictions of the proposed ILP formulation, present several classes of facet-defining inequalities and discuss separation algorithms.     

Models and algorithms for skip-free Markov decision processes on trees

We introduce a class of models for multidimensional control problems that we call skip-free Markov decision processes on trees. We describe and analyse an algorithm applicable to Markov decision processes of this type that are skip-free in the negative direction. Starting with the finite average cost case, we show that the algorithm combines the advantages of both value iteration and policy iteration—it is guaranteed to converge to an optimal policy and optimal value function after a finite number of iterations but the computational effort required for each iteration step is comparable with that for value iteration. We show that the algorithm can also be used to solve discounted cost models and continuous-time models, and that a suitably modified algorithm can be used to solve communicating models.     

Metropolis–Hastings Algorithms with acceptance ratios of nearly 1

We develop the results on polynomial ergodicity of Markov chains and apply to the Metropolis–Hastings algorithms based on a Langevin diffusion. When a prescribed distribution p has heavy tails, the Metropolis–Hastings algorithms based on a Langevin diffusion do not converge to p at any geometric rate. However, those Langevin based algorithms behave like the diffusion itself in the tail area, and using this fact, we provide sufficient conditions of a polynomial rate convergence. By the feature in the tail area, our results can be applied to a large class of distributions to which p belongs. Then, we show that the convergence rate can be improved by a transformation. We also prove central limit theorems for those algorithms.     

Discounted Markov games: Generalized policy iteration method

In this paper, we consider two-person zero-sum discounted Markov games with finite state and action spaces. We show that the Newton-Raphson or policy iteration method as presented by Pollats-chek and Avi-Itzhak does not necessarily converge, contradicting a proof of Rao, Chandrasekaran, and Nair. Moreover, a set of successive approximation algorithms is presented of which Shapley's method and a total-expected-rewards version of Hoffman and Karp's method are the extreme elements.     

Backward–forward algorithms for structured monotone inclusions in Hilbert spaces

In this paper, we study the backward–forward algorithm as a splitting method to solve structured monotone inclusions, and convex minimization problems in Hilbert spaces. It has a natural link with the forward–backward algorithm and has the same computational complexity, since it involves the same basic blocks, but organized differently. Surprisingly enough, this kind of iteration arises when studying the time discretization of the regularized Newton method for maximally monotone operators. First, we show that these two methods enjoy remarkable involutive relations, which go far beyond the evident inversion of the order in which the forward and backward steps are applied. Next, we establish several convergence properties for both methods, some of which were unknown even for the forward–backward algorithm. This brings further insight into this well-known scheme. Finally, we specialize our results to structured convex minimization problems, the gradient-projection algorithms, and give a numerical illustration of theoretical interest.     

Markov Incremental Constructions

A classic result asserts that many geometric structures can be constructed optimally by successively inserting their constituent parts in random order. These randomized incremental constructions (RICs) still work with imperfect randomness: the dynamic operations need only be “locally” random. Much attention has been given recently to inputs generated by Markov sources. These are particularly interesting to study in the framework of RICs, because Markov chains provide highly nonlocal randomness, which incapacitates virtually all known RIC technology. We generalize Mulmuley’s theory of Θ-series and prove that Markov incremental constructions with bounded spectral gap are optimal within polylog factors for trapezoidal maps, segment intersections, and convex hulls in any fixed dimension. The main contribution of this work is threefold: (i) extending the theory of abstract configuration spaces to the Markov setting; (ii) proving Clarkson–Shor-type bounds for this new model; (iii) applying the results to classical geometric problems. We hope that this work will pioneer a new approach to randomized analysis in computational geometry. This work was supported in part by NSF grants CCR-0306283, CCF-0634958.     

Boolean autoencoders and hypercube clustering complexity

We introduce and study the properties of Boolean autoencoder circuits. In particular, we show that the Boolean autoencoder circuit problem is equivalent to a clustering problem on the hypercube. We show that clustering m binary vectors on the n-dimensional hypercube into k clusters is NP-hard, as soon as the number of clusters scales like ${m^\epsilon (\epsilon >0 )}$ , and thus the general Boolean autoencoder problem is also NP-hard. We prove that the linear Boolean autoencoder circuit problem is also NP-hard, and so are several related problems such as: subspace identification over finite fields, linear regression over finite fields, even/odd set intersections, and parity circuits. The emerging picture is that autoencoder optimization is NP-hard in the general case, with a few notable exceptions including the linear cases over infinite fields or the Boolean case with fixed size hidden layer. However learning can be tackled by approximate algorithms, including alternate optimization, suggesting a new class of learning algorithms for deep networks, including deep networks of threshold gates or artificial neurons.     

On Assessing the Performance of Randomized Algorithms

In this paper we study randomized algorithms with random input. We adapt to such algorithms the notion of probability of a false positive which is common in epidemiological studies. The probability of a false positive takes into account both the (controlled) error of the randomization and the randomness of the input, which needs to be modeled. We illustrate our idea on two classes of problems: primality testing and fingerprinting in strings transmission. Although in both cases the randomization has low error, in the first one the probability of a false positive is very low, while in the second one it is not. We end the paper with a discussion of randomness illustrated in a textbook example.     

Random Block Coordinate Descent Methods for Linearly Constrained Optimization over Networks

In this paper we develop random block coordinate descent methods for minimizing large-scale linearly constrained convex problems over networks. Since coupled constraints appear in the problem, we devise an algorithm that updates in parallel at each iteration at least two random components of the solution, chosen according to a given probability distribution. Those computations can be performed in a distributed fashion according to the structure of the network. Complexity per iteration of the proposed methods is usually cheaper than that of the full gradient method when the number of nodes in the network is much larger than the number of updated components. On smooth convex problems, we prove that these methods exhibit a sublinear worst-case convergence rate in the expected value of the objective function. Moreover, this convergence rate depends linearly on the number of components to be updated. On smooth strongly convex problems we prove that our methods converge linearly. We also focus on how to choose the probabilities to make our randomized algorithms converge as fast as possible, which leads us to solving a sparse semidefinite program. We then describe several applications that fit in our framework, in particular the convex feasibility problem. Finally, numerical experiments illustrate the behaviour of our methods, showing in particular that updating more than two components in parallel accelerates the method.     

Progressive Regularization of Variational Inequalities and Decomposition Algorithms

For nonsymmetric operators involved in variational inequalities, the strong monotonicity of their possibly multivalued inverse operators (referred to as the Dunn property) appears to be the weakest requirement to ensure convergence of most iterative algorithms of resolution proposed in the literature. This implies the Lipschitz property, and both properties are equivalent for symmetric operators. For Lipschitz operators, the Dunn property is weaker than strong monotonicity, but is stronger than simple monotonicity. Moreover, it is always enforced by the Moreau–Yosida regularization and it is satisfied by the resolvents of monotone operators. Therefore, algorithms should always be applied to this regularized version or they should use resolvents: in a sense, this is what is achieved in proximal and splitting methods among others. However, the operation of regularization itself or the computation of resolvents may be as complex as solving the original variational inequality. In this paper, the concept of progressive regularization is introduced and a convergent algorithm is proposed for solving variational inequalities involving nonsymmetric monotone operators. Essentially, the idea is to use the auxiliary problem principle to perform the regularization operation and, at the same time, to solve the variational inequality in its approximately regularized version; thus, two iteration processes are performed simultaneously, instead of being nested in each other, yielding a global explicit iterative scheme. Parallel and sequential versions of the algorithm are presented. A simple numerical example demonstrates the behavior of these two versions for the case where previously proposed algorithms fail to converge unless regularization or computation of a resolvent is performed at each iteration. Since the auxiliary problem principle is a general framework to obtain decomposition methods, the results presented here extend the class of problems for which decomposition methods can be used.     

Greedy expansions in convex optimization

This paper is a follow-up to the author’s previous paper on convex optimization. In that paper we began the process of adjusting greedy-type algorithms from nonlinear approximation for finding sparse solutions of convex optimization problems. We modified there the three most popular greedy algorithms in nonlinear approximation in Banach spaces-Weak Chebyshev Greedy Algorithm, Weak Greedy Algorithm with Free Relaxation, and Weak Relaxed Greedy Algorithm-for solving convex optimization problems. We continue to study sparse approximate solutions to convex optimization problems. It is known that in many engineering applications researchers are interested in an approximate solution of an optimization problem as a linear combination of elements from a given system of elements. There is an increasing interest in building such sparse approximate solutions using different greedy-type algorithms. In this paper we concentrate on greedy algorithms that provide expansions, which means that the approximant at the mth iteration is equal to the sum of the approximant from the previous, (m ? 1)th, iteration and one element from the dictionary with an appropriate coefficient. The problem of greedy expansions of elements of a Banach space is well studied in nonlinear approximation theory. At first glance the setting of a problem of expansion of a given element and the setting of the problem of expansion in an optimization problem are very different. However, it turns out that the same technique can be used for solving both problems. We show how the technique developed in nonlinear approximation theory, in particular, the greedy expansions technique, can be adjusted for finding a sparse solution of an optimization problem given by an expansion with respect to a given dictionary.     

Randomized Algorithms over Finite Fields for the Exact Parity Base Problem

We present three randomized pseudo-polynomial algorithms for the problem of finding a base of specified value in a weighted represented matroid subject to parity conditions. These algorithms, the first two being an improved version of those presented by P. M. Camerini et al. (1992, J. Algorithms13, 258–273) use fast arithmetic working over a finite field chosen at random among a set of appropriate fields. We show that the choice of a best algorithm among those presented depends on a conjecture related to the best value of the so-called Linnik constant concerning the distribution of prime numbers in arithmetic progressions. This conjecture, which we call the C-conjecture, is a strengthened version of a conjecture formulated in 1934 by S. Chowla. If the C-conjecture is true, the choice of a best algorithm is simple, since the last algorithm exhibits the best performance, either when the performance is measured in arithmetic operations, or when it is measured in bit operations and mild assumptions hold. If the C-conjecture is false we are still able to identify a best algorithm, but in this case the choice is between the first two algorithms and depends on the asymptotic growth of m with respect to those of U and n, where 2n, 2m, U are the rank, the number of elements, and the maximum weight assigned to the elements of the matroid, respectively.     

A family of projective splitting methods for the sum of two maximal monotone operators

A splitting method for two monotone operators A and B is an algorithm that attempts to converge to a zero of the sum A + B by solving a sequence of subproblems, each of which involves only the operator A, or only the operator B. Prior algorithms of this type can all in essence be categorized into three main classes, the Douglas/Peaceman-Rachford class, the forward-backward class, and the little-used double-backward class. Through a certain “extended” solution set in a product space, we construct a fundamentally new class of splitting methods for pairs of general maximal monotone operators in Hilbert space. Our algorithms are essentially standard projection methods, using splitting decomposition to construct separators. We prove convergence through Fejér monotonicity techniques, but showing Fejér convergence of a different sequence to a different set than in earlier splitting methods. Our projective algorithms converge under more general conditions than prior splitting methods, allowing the proximal parameter to vary from iteration to iteration, and even from operator to operator, while retaining convergence for essentially arbitrary pairs of operators. The new projective splitting class also contains noteworthy preexisting methods either as conventional special cases or excluded boundary cases. Dedicated to Clovis Gonzaga on the occassion of his 60th birthday.     

Accelerated PMHSS iteration methods for complex symmetric linear systems

In this paper, we introduce and analyze an accelerated preconditioning modification of the Hermitian and skew-Hermitian splitting (APMHSS) iteration method for solving a broad class of complex symmetric linear systems. This accelerated PMHSS algorithm involves two iteration parameters α,β and two preconditioned matrices whose special choices can recover the known PMHSS (preconditioned modification of the Hermitian and skew-Hermitian splitting) iteration method which includes the MHSS method, as well as yield new ones. The convergence theory of this class of APMHSS iteration methods is established under suitable conditions. Each iteration of this method requires the solution of two linear systems with real symmetric positive definite coefficient matrices. Theoretical analyses show that the upper bound σ₁(α,β) of the asymptotic convergence rate of the APMHSS method is smaller than that of the PMHSS iteration method. This implies that the APMHSS method may converge faster than the PMHSS method. Numerical experiments on a few model problems are presented to illustrate the theoretical results and examine the numerical effectiveness of the new method.