On the complexity of a two-point boundary value problem in different settings

We study the complexity of a two-point boundary value problem. We concentrate on the linear problem of order k with separated boundary conditions. Right-hand side functions are assumed to be r times differentiable with all derivatives bounded by a constant. We consider three models of computation: deterministic with standard and linear information, randomized and quantum. In each setting, we construct an algorithm for solving the problem, which allows us to establish upper complexity bounds. In the deterministic setting, we show that the use of linear information gives us a speed-up of at least one order of magnitude compared with the standard information. For randomized algorithms, we show that the speed-up over standard deterministic algorithms is by 1/2 in the exponent. For quantum algorithms, we can achieve a speed-up by one order of magnitude. We also provide lower complexity bounds. They match upper bounds in the deterministic setting with the standard information, and almost match upper bounds in the randomized and quantum settings. In the deterministic setting with the linear information, a gap still remains between the upper and lower complexity bounds.     

Learning with continuous experts using drifting games

We consider the problem of learning to predict as well as the best in a group of experts making continuous predictions. We assume the learning algorithm has prior knowledge of the maximum number of mistakes of the best expert. We propose a new master strategy that achieves the best known performance for on-line learning with continuous experts in the mistake bounded model. Our ideas are based on drifting games, a generalization of boosting and on-line learning algorithms. We prove new lower bounds based on the drifting games framework which, though not as tight as previous bounds, have simpler proofs and do not require an enormous number of experts. We also extend previous lower bounds to show that our upper bounds are exactly tight for sufficiently many experts. A surprising consequence of our work is that continuous experts are only as powerful as experts making binary or no prediction in each round.     

Minimum-diameter cyclic arrangements in mapping data-flow graphs onto VLSI arrays

Regular arrays of processing elements in VLSI have proved to be suitable for high-speed execution of many matrix operations. To execute an arbitrary computational algorithm on such processing arrays, it has been suggested mapping the given algorithm directly onto a regular array. The computational algorithm is represented by a data-flow graph whose nodes are to be mapped onto processors in the VLSI array. This study examines the complexity of mapping data-flow graphs onto square and hexagonal arrays of processors. We specifically consider the problem of routing data from processors in a given (source) sequence to another (target) sequence. We show that under certain conditions, the above problem is equivalent to the one of finding a minimum-diameter cyclic arrangement. The complexity of the latter problem is analyzed and upper and lower bounds on the number of intermediate rows of processors (between the source and target rows) are derived.     

Resolution for Max-SAT

Max-SAT is the problem of finding an assignment minimizing the number of unsatisfied clauses in a CNF formula. We propose a resolution-like calculus for Max-SAT and prove its soundness and completeness. We also prove the completeness of some refinements of this calculus. From the completeness proof we derive an exact algorithm for Max-SAT and a time upper bound.We also define a weighted Max-SAT resolution-like rule, and show how to adapt the soundness and completeness proofs of the Max-SAT rule to the weighted Max-SAT rule.Finally, we give several particular Max-SAT problems that require an exponential number of steps of our Max-SAT rule to obtain the minimal number of unsatisfied clauses of the combinatorial principle. These results are based on the corresponding resolution lower bounds for those particular problems.     

A two-machine flowshop problem with processing time-dependent buffer constraints—An application in multimedia presentations

To have a quality multimedia presentation through networks, its presentation lag needs to be controlled. One way to reduce the lag is to prefetch the media objects before their due dates. This paper explores techniques for optimizing the object sequence in a prefetch-enabled TV-like presentation. An optimal solution is the one with which the presentation lag is minimized. We formulate the problem into a two-machine flowshop scheduling problem with a single chain precedence constraint and a player-side buffer constraint. The player-side buffer is “processing time-dependent” and distinguished from the conventional item-based intermediate buffer constraints discussed in previous flowshop studies. We prove the problem to be strongly NP-hard. A branch and bound algorithm equipped with four lower bounds and an NEH-based upper bound is developed. The simulation results show that the average gaps between the overall lower bounds and the NEH-based upper bound are less than 3% for problems with a large buffer size, and less than 13% for problems with a small buffer size and high density of precedence constraints. For applications where the media objects are delivered through extremely busy servers with which only very restricted CPU resources can be allocated for computation, the CDS-based algorithm provides better sequences than the NEH-based algorithm.     

Branch and bound for the cutwidth minimization problem

The cutwidth minimization problem consists of finding a linear arrangement of the vertices of a graph where the maximum number of cuts between the edges of the graph and a line separating consecutive vertices is minimized. We first review previous approaches for special classes of graphs, followed by lower bounds and then a linear integer formulation for the general problem. We then propose a branch-and-bound algorithm based on different lower bounds on the cutwidth of partial solutions. Additionally, we introduce a Greedy Randomized Adaptive Search Procedure (GRASP) heuristic to obtain good initial solutions. The combination of the branch-and-bound and GRASP methods results in optimal solutions or a reduced relative gap (difference between upper and lower bounds) on the instances tested. Empirical results with a collection of previously reported instances indicate that the proposed algorithm is able to solve all the small instances (up to 32 vertices) as well as some of the large instances tested (up to 158 vertices) using less than 30 minutes of CPU time. We compare the results of our method with previous lower bounds, and with the best previous linear integer formulation solved using Cplex. Both comparisons favor the proposed procedure.     

On-Line Algorithms for the Dynamic Traveling Repair Problem

We consider the dynamic traveling repair problem in which requests with deadlines arrive through time on points in a metric space. Servers move from point to point at constant speed. The goal is to plan the motion of servers so that the maximum number of requests are met by their deadline. We consider a restricted version of the problem in which there is a single server and the length of time between the arrival of a request and its deadline is constant. We give upper bounds for the competitive ratio of two very natural algorithms as well as several lower bounds for any deterministic algorithm. Most of the results in this paper are expressed as a function of β, the diameter of the metric space. In particular, we prove that the upper bound given for one of the two algorithms is within a constant factor of the best possible competitive ratio.     

On the Worst-Case Analysis of Temporal-Difference Learning Algorithms

We study the behavior of a family of learning algorithms based on Suttons method of temporal differences. In our on-line learning framework, learning takes place in a sequence of trials, and the goal of the learning algorithm is to estimate a discounted sum of all the reinforcements that will be received in the future. In this setting, we are able to prove general upper bounds on the performance of a slightly modified version of Suttons so-called TD((gl) algorithm. These bounds are stated in terms of the performance of the best linear predictor on the given training sequence, and are proved without making any statistical assumptions of any kind about the process producing the learners observed training sequence. We also prove lower bounds on the performance of any algorithm for this learning problem, and give a similar analysis of the closely related problem of learning to predict in a model in which the learner must produce predictions for a whole batch of observations before receiving reinforcement.     

Kolmogorov complexity and combinatorial methods in communication complexity

We introduce a method based on Kolmogorov complexity to prove lower bounds on communication complexity. The intuition behind our technique is close to information theoretic methods.We use Kolmogorov complexity for three different things: first, to give a general lower bound in terms of Kolmogorov mutual information; second, to prove an alternative to Yao’s minmax principle based on Kolmogorov complexity; and finally, to identify hard inputs.We show that our method implies the rectangle and corruption bounds, known to be closely related to the subdistribution bound. We apply our method to the hidden matching problem, a relation introduced to prove an exponential gap between quantum and classical communication. We then show that our method generalizes the VC dimension and shatter coefficient lower bounds. Finally, we compare one-way communication and simultaneous communication in the case of distributional communication complexity and improve the previous known result.     

Selection problems viaM-ary queries

It is well known that, for fixedk, to find thek-th largest ofn elementsn+(k?1)log₂ n+Θ(1) comparisons are necessary and sufficient. But do the same bounds apply if we use a different type of query? We show that the arity of the queries is relevant. In particular, we present upper and lower bounds for finding the maximum using 3-ary or 4-ary Boolean (YES/NO answers) queries. We also study general (e.g.,max, sort) 3-ary queries, and show bounds for finding the maximum and the second largest. For sort queries we show matching upper and lower bounds.     

Randomized competitive algorithms for the list update problem

We prove upper and lower bounds on the competitiveness of randomized algorithms for the list update problem of Sleator and Tarjan. We give a simple and elegant randomized algorithm that is more competitive than the best previous randomized algorithm due to Irani. Our algorithm uses randomness only during an initialization phase, and from then on runs completely deterministically. It is the first randomized competitive algorithm with this property to beat the deterministic lower bound. We generalize our approach to a model in which access costs are fixed but update costs are scaled by an arbitrary constantd. We prove lower bounds for deterministic list update algorithms and for randomized algorithms against oblivious and adaptive on-line adversaries. In particular, we show that for this problem adaptive on-line and adaptive off-line adversaries are equally powerful.A preliminary version of these results appeared in a joint paper with S. Irani in theProceedings of the 2nd Symposium on Discrete Algorithms, 1991 [17].This research was partially supported by NSF Grants CCR-8808949 and CCR-8958528.This research was partially supported by NSF Grant CCR-9009753.This research was supported in part by the National Science Foundation under Grant CCR-8658139, by DIMACS, a National Science Foundation Science and Technology center, Grant No. NSF-STC88-09648.     

Improved upper bounds for the mixed structured singular value

In this paper, we take a new look at the mixed structured singular value problem, a problem of finding important applications in robust stability analysis. Several new upper bounds are proposed using a very simple approach which we call the multiplier approach. These new bounds are convex and computable by using linear matrix inequality (LMI) techniques. We show, most importantly, that these upper bounds are actually lower bounds of a well-known upper bound which involves the so-called D-scaling (for complex perturbations) and G-scaling (for real perturbations)     

Hitting time of quantum walks with perturbation

The hitting time is the required minimum time for a Markov chain-based walk (classical or quantum) to reach a target state in the state space. We investigate the effect of the perturbation on the hitting time of a quantum walk. We obtain an upper bound for the perturbed quantum walk hitting time by applying Szegedy’s work and the perturbation bounds with Weyl’s perturbation theorem on classical matrix. Based on the definition of quantum hitting time given in MNRS algorithm, we further compute the delayed perturbed hitting time and delayed perturbed quantum hitting time (DPQHT). We show that the upper bound for DPQHT is bounded from above by the difference between the square root of the upper bound for a perturbed random walk and the square root of the lower bound for a random walk.     

Simple On-Line Algorithms for the Maximum Disjoint Paths Problem

In this paper we study the classical problem of finding disjoint paths in graphs. This problem has been studied by a number of authors both for specific graphs and general classes of graphs. Whereas for specific graphs many (almost) matching upper and lower bounds are known for the competitiveness of on-line algorithms, not much is known about how well on-line algorithms can perform in the general setting. The best results obtained so far use the expansion of a network to measure the algorithms performance. We use a different parameter called the routing number that, as we will show, allows more precise results than the expansion. It enables us to prove tight upper and lower bounds for deterministic on-line algorithms. The upper bound is obtained by surprisingly simple greedy-like algorithms. Interestingly, our upper bound on the competitive ratio is even better than the best previous approximation ratio for off-line algorithms. Furthermore, we introduce a refined variant of the routing number and show that this variant allows us, for some classes of graphs, to construct on-line algorithms with a competitive ratio significantly below the best possible upper bound that could be obtained using the routing number or the expansion of a network only. We also show that our on-line algorithms can be transformed into efficient algorithms for the more general unsplittable flow problem.     

On the complexity of finding a local maximum of functions on discrete planar subsets

We study how many values of an unknown integer-valued function f one needs to know in order to find a local maximum of f. We consider functions defined on finite subsets of discrete plane. We prove upper bounds for functions defined on rectangles and present lower bounds for functions defined on arbitrary domains in terms of the size of the domain and the size of its border.     

New closed-form bounds on the partition function

Estimating the partition function is a key but difficult computation in graphical models. One approach is to estimate tractable upper and lower bounds. The piecewise upper bound of Sutton et al. is computed by breaking the graphical model into pieces and approximating the partition function as a product of local normalizing factors for these pieces. The tree reweighted belief propagation algorithm (TRW-BP) by Wainwright et al. gives tighter upper bounds. It optimizes an upper bound expressed in terms of convex combinations of spanning trees of the graph. Recently, Globerson et al. gave a different, convergent iterative dual optimization algorithm TRW-GP for the TRW objective. However, in many practical applications, particularly those that train CRFs with many nodes, TRW-BP and TRW-GP are too slow to be practical. Without changing the algorithm, we prove that TRW-BP converges in a single iteration for associative potentials, and give a closed form for the solution it finds. The closed-form solution obviates the need for complex optimization. We use this result to develop new closed-form upper bounds for MRFs with arbitrary pairwise potentials. Being closed-form, they are much faster to compute than TRW-based bounds. We also prove similar convergence results for loopy belief propagation (LBP) and use it to obtain closed-form solutions to the LBP pseudomarginals and approximation to the partition function for associative potentials. We then use recent results proved by Wainwright et al for binary MRFs to obtain closed-form lower bounds on the partition function. We then develop novel lower bounds for arbitrary associative networks. We report on experiments with synthetic and real-world graphs. Our new upper bounds are considerably tighter than the piecewise bounds in practice. Moreover, we can compute our bounds on several graphs where TRW-BP does not converge. Our novel lower bound, in spite of being closed-form and much faster to compute, outperforms more complicated popular algorithms for computing lower bounds like mean-field on densely connected graphs by wide margins although it does worse on sparsely connected graphs like chains.     

A note on monotone complexity and the rank of matrices

We shall give simpler proofs of some lower bounds on monotone computations. We describe a simple condition on combinatorial structures, such that the rank of the matrix associated with these structures gives lower bounds on monotone span program size and monotone formula size. We also prove an upper bound on the rank of the corresponding matrices, and show that such structures can be constructed from self-avoiding families. As a corollary, we obtain an upper bound on the size of self-avoiding families, which solves a problem posed by Babai and Gál [Combinatorica 19 (3) (1999) 301-319].     

Classifying the multi robot path finding problem into a quadratic competitive complexity class

We explore two motion planning problems where a group of mobile robots has to reach a target located in an a priori unknown environment while on-line planning the next step. In the first problem the target position is unknown and should be found by the robots, while in the second problem the target position is known and only a path to it should be found. We focus on optimizing the cost of the task in terms of motion time, which, under the assumption of uniform velocity of all the robots, correlates to the path length passed by the robot which reaches the target. The performance of an on-line algorithm is usually expressed in terms of Competitiveness, the constant ratio between the on-line and the optimal off-line solutions. Specifically, the ratio between the lengths of the actual path made by the robot which reached the target to the shortest path to the target. We use generalized competitiveness, i.e., the ratio is not necessarily constant, but could be any function. Classification of a motion planning task in the sense of performance is done by finding an upper and a lower bounds on the competitiveness of all algorithms solving that task. If the two bounds belong to the same functional class this is the Competitive Complexity Class of the task. We find the two bounds for the aforementioned common on-line motion planning problems, and classify them into competitive classes. It is shown that in general any on-line motion planning algorithm that tries to solve these problems must have at least a quadratic competitive performance. This is a lower bound of the problems. This paper describes two new on-line navigation algorithm which solve the problems under discussion. The first is called MRSAM, short for Multi-Robot Search Area Multiplication, and the second is called MRBUG, short for Multi-Robot BUG which extends Lumelsky famous BUG algorithm. Both algorithms have quadratic upper bounds, which prove that the problems they solve have quadratic upper bounds. Thus it is shown that navigation in an unknown environment by a group of robots belongs to a quadratic competitive class. MRSAM and MRBUG have a quadratic competitive performance and thus have optimal competitiveness. The algorithms’ performance is simulated in office-like environments.     

Safe bounds for the solutions of nonlinear problems using a parallel multisplitting method

For some systems of nonlinear equationsF(x)=0 we derive an algorithm which iteratively constructs tight lower and upper bounds for the zeros ofF. The algorithm is based on a multisplitting of certain matrices thus showing a natural parallelism. We prove criteria for the convergence of the bounds towards the zeros and we investigate the speed of convergence.     

On the Role of Hadamard Gates in Quantum Circuits

We study a reduced quantum circuit computation paradigm in which the only allowable gates either permute the computational basis states or else apply a “global Hadamard operation”, i.e. apply a Hadamard operation to every qubit simultaneously. In this model, we discuss complexity bounds (lower-bounding the number of global Hadamard operations) for common quantum algorithms: we illustrate upper bounds for Shor’s Algorithm, and prove lower bounds for Grover’s Algorithm. We also use our formalism to display a gate that is neither quantum-universal nor classically simulable, on the assumption that Integer Factoring is not in BPP.