In this article we present a theoretical analysis of the online Sum-of-Squares algorithm (SS) for bin packing along with several new variants. SS is applicable to any instance of bin packing in which the bin capacity B and item sizes s(a) are integral (or can be scaled to be so), and runs in time O(nB). It performs remarkably well from an average case point of view: For any discrete distribution in which the optimal expected waste is sublinear, SS also has sublinear expected waste. For any discrete distribution where the optimal expected waste is bounded, SS has expected waste at most O(log n). We also discuss several interesting variants on SS, including a randomized O(nBlog B)-time online algorithm SS ∗ whose expected behavior is essentially optimal for all discrete distributions. Algorithm SS ∗ depends on a new linear-programming-based pseudopolynomial-time algorithm for solving the

We consider packet routing when packets are injected continuously into a network. We develop an adversarial theory of queuing aimed at addressing some of the restrictions inherent in probabilistic analysis and queuing theory based on time-invariant stochastic generation. We examine the stability of queuing networks and policies when the arrival process is adversarial, and provide some preliminary results in this direction. Our approach sheds light on various queuing policies in simple networks, and paves the way for a systematic study of queuing with few or no probabilistic assumptions.

Randomized search heuristics like evolutionary algorithms are mostly applied to problems whose structure is not completely known but also to combinatorial optimization problems. Practitioners report surprising successes but almost no results with theoretically well-founded analyses exist. Such an analysis is started in this paper for a fundamental evolutionary algorithm and the well-known maximum matching problem. It is

In recent years, gossip-based algorithms have gained prominence as a methodology for designing robust and scalable communication schemes in large distributed systems. The premise underlying distributed gossip is very simple: in each time step, each node v in the system selects some other node w as a communication partner — generally by a simple randomized rule — and exchanges information with w; over a period of time, information spreads through the system in an “epidemic fashion”. A fundamental issue which is not well understood is the following: how does the underlying low-level gossip mechanism — the means by which communication partners are chosen — affect one’s ability to design efficient high-level gossip-based protocols? We establish one of the first concrete results addressing this question, by showing a fundamental limitation on the power of the commonly used uniform gossip mechanism for solving nearest-resource location problems. In contrast, very efficient protocols for this problem can be designed using a non-uniform spatial gossip mechanism, as established in earlier work with Alan Demers. We go on to consider the design of protocols for more complex problems, providing an efficient distributed gossipbased protocol for a set of nodes in Euclidean space to construct an approximate minimum spanning tree. Here too, we establish a contrasting limitation on the power of uniform gossip for solving this problem. Finally, we investigate gossip-based packet routing as a primitive that underpins the communication patterns in many protocols, and as a way to understand the capabilities of different gossip mechanisms at a general level.

In this paper, we analyze the behavior of packet-switched communication networks in which packets arrive dynamically at the nodes and are routed in discrete time steps across the edges. We focus on a basic adversarial model of packet arrival and path determination for which the time--averaged arrival rate of packets requiring the use of any edge is limited to be less than 1. This model can reflect the behavior of connection-oriented networks with transient connections (such as ATM networks) as well as connectionless networks (such as the Internet). We concentrate on

We study the problem of dynamic routing on arrays. We prove that a large class of greedy algorithms perform very well on average. In the dynamic case, when the arrival rate of packets in an N \Theta N array is at most 99% of network capacity, we establish an exponential bound on the tail of the delay distribution. Moreover, we show that, in any window of T steps, the maximum queue-size is O(1 + log T= log N ) with high probability. We extend these results to the case of bit-serial routing, and to the static case. We also calculate the exact value of the ergodic expected delay and queue-sizes under the farthest-first protocol for the one dimensional array, and for the ring when the arrivals are Poisson. 1 Introduction Many parallel machines, such as the MPP, Ametek and Intel Touchstone are configured as a low-dimensional array containing a large number of processors. These machines generally route packets using simple greedy algorithms. While these algorithms tend to behave well experi...

We study the Best Fit algorithm for on-line bin packing under the distribution in which the item sizes are uniformly distributed in the discrete range f1=k � 2=k�:::�j=kg. Our main result is that, in the case j = k; 2, the asymptotic expected waste remains bounded. This settles an open problem of Co man et al [3], and involves a detailed analysis of the in nite multi-dimensional Markov chain underlying the algorithm.

Randomized search heuristics like evolutionary algorithms and simulated annealing find many applications, especially in situations where no full information on the problem instance is available. In order to understand how these heuristics work, it is necessary to analyze their behavior on classes of functions. Such an analysis is performed here for the class of monotone pseudo-boolean polynomials. Results depending on the degree and the number of terms of the polynomial are obtained. The class of monotone polynomials is of special interest since simple functions of this kind can have an image set of exponential size, improvements can increase the Hamming distance to the optimum and in order to find a better search point, it can be necessary to search within a large plateau of search points with the same fitness value.

Using stochastic dominance, in this paper we provide a new characterization of point processes. This characterization leads to a unified proof for various stability results of open Jackson networks where service times are i.i.d. with a general distribution, external interarrival times are i.i.d. with a general distribution and the routing is Bernoulli. We show that if the traffic condition is satisfied, i.e., the input rate is smaller than the service rate at each queue, then the queue length process (the number of customers at each queue) is tight. Under the traffic condition, the p th moment of the queue length process is bounded for all t if the p+1 th moment of the service times at all queues are nite. If, furthermore, the moment generating functions of the service times at all queues exist, then all the moments of the queue length process are bounded for all t. When the interarrival times are unbounded and non-lattice (resp. spread-out), the queue lengths and the remaining service times converge in distribution (resp. in total variation) to a steady state. Also, the moments converge if the corresponding moment conditions are satisfied.

Abstract. Multi-objective evolutionary algorithms (MOEAs) have become increasingly popular as multi-objective problem solving techniques. Most studies of MOEAs are empirical. Only recently, a few theoretical results have appeared. It is acknowledged that more theoretical research is needed. An important open problem is to understand the role of populations in MOEAs. We present a simple bi-objective problem which emphasizes when populations are needed. Rigorous runtime analysis point out an exponential runtime gap between a population-based algorithm (SEMO) and several single individual-based algorithms on this problem. This means that among the algorithms considered, only the populationbased MOEA is successful and all other algorithms fail. 1