Ronald L. Rardin

Experimental Evaluation of Heuristic Optimization Algorithms: A Tutorial

Heuristic optimization algorithms seek good feasible solutions to optimization problems in circumstances where the complexities of the problem or the limited time available for solution do not allow exact solution. Although worst case and probabilistic analysis of algorithms have produced insight on some classic models, most of the heuristics developed for large optimization problem must be evaluated empirically—by applying procedures to a collection of specific instances and comparing the observed solution quality and computational burden.This paper focuses on the methodological issues that must be confronted by researchers undertaking such experimental evaluations of heuristics, including experimental design, sources of test instances, measures of algorithmic performance, analysis of results, and presentation in papers and talks. The questions are difficult, and there are no clear right answers. We seek only to highlight the main issues, present alternative ways of addressing them under different circumstances, and caution about pitfalls to avoid.

Tractable nonlinear production planning models for semiconductor wafer fabrication facilities

We describe a simulation study of a production planning model for multistage production inventory systems that reflects the nonlinear relationship between resource utilization and lead time. The model is based on the use of clearing functions that capture the nonlinear relationship between workload and throughput. We show how these clearing functions can be estimated from empirical data using a simulation model as a surrogate for observation of the production system under study. We then examine the sensitivity of the estimated clearing function to different dispatching algorithms, different demand patterns, and production planning techniques. Computational experiments based on a scaled-down model of a semiconductor wafer fabrication facility illustrate the potential benefits of the clearing function model relative to conventional linear programming models.

Matching daily healthcare provider capacity to demand in advanced access scheduling systems

Advanced access scheduling, introduced in the early 1990s, is reported to significantly improve the performance of outpatient clinics. The successful implementation of advanced access scheduling requires the match of daily healthcare provider capacity with patient demand. In this paper, for the first time a closed-form approach is presented to determine the optimal percentage of open-access appointments to match daily provider capacity to demand. This paper introduces the conditions for the optimal percentage of open-access appointments and the procedure to find the optimal percentage. Furthermore, the sensitivity of the optimal percentage of open-access appointments to provider capacity, no-show rates, and demand distribution is investigated. Our results demonstrate that the optimal percentage of open-access appointments mainly depends on the ratio of the average demand for open-access appointments to provider capacity and the ratio of the show-up rates for prescheduled and open-access appointments.

Some relationships between lagrangian and surrogate duality in integer programming

Lagrangian dual approaches have been employed successfully in a number of integer programming situations to provide bounds for branch-and-bound procedures. This paper investigates some relationship between bounds obtained from lagrangian duals and those derived from the lesser known, but theoretically more powerful surrogate duals. A generalization of Geoffrions integrality property, some complementary slackness relationships between optimal solutions, and some empirical results are presented and used to argue for the relative value of surrogate duals in integer programming. These and other results are then shown to lead naturally to a two-phase algorithm which optimizes first the computationally easier lagrangian dual and then the surrogate dual.

A coupled column generation, mixed integer approach to optimal planning of intensity modulated radiation therapy for cancer

Abstract.Approximately 40% of all U.S. cancer cases are treated with radiation therapy. In Intensity-Modulated Radiation Therapy (IMRT) the treatment planning problem is to choose external beam angles and their corresponding intensity maps (showing how the intensity varies across a given beam) to maximize tumor dose subject to the tolerances of surrounding healthy tissues. Dose, like temperature, is a quantity defined at each point in the body, and the distribution of dose is determined by the choice of treatment parameters available to the planner. In addition to absolute dose limits in healthy tissues, some tissues have at least one dose-volume restriction that requires a fraction of its volume to not exceed a specified tighter threshold level for damage. There may also be a homogeneity limit for the tumor that restricts the allowed spread of dose across its volume. We formulate this planning problem as a mixed integer program over a coupled pair of column generation processes -- one designed to produce intensity maps, and a second specifying protected area choices for tissues under dose-volume restrictions. The combined procedure is shown to strike a balance between computing an approximately optimal solution and bounding its maximum possible suboptimality that we believe holds promise for implementations able to offer the power and flexibility of mixed-integer linear programming models on instances of practical scale.

Polyhedral characterization of discrete dynamic programming

Many interesting combinatorial problems can be optimized efficiently using recursive computations often termed discrete dynamic programming. In this paper, we develop a paradigm for a general class of such optimizations that yields a polyhedral description for each model in the class. The elementary concept of dynamic programs as shortest path problems in acyclic graphs is generalized to one seeking a least cost solution in a directed hypergraph. Sufficient conditions are then provided for binary integrality of the associated hyperflow problem. Given a polynomially solvable dynamic program, the result is a linear program, in polynomially many variables and constraints, having the solution vectors of the dynamic program as its extreme-point optima. That is, the linear program provides a succinct characterization of the solutions to the underlying optimization problem expressed through an appropriate change of variables. We also discuss projecting this formulation to recover constraints on the original variables and illustrate how some important dynamic programming solvable models fit easily into our paradigm. A classic multiechelon lot sizing problem in production and a host of optimization problems on recursively defined classes of graphs are included.

Valid inequalities and projecting the multicommodity extended formulation for uncapacitated fixed charge network flow problems

Multicommodity extended formulations of single source uncapacitated fixed charge network flow problems have significantly sharper linear programming relaxations than the standard flow formulations. However the tradeoff is the introduction of many new constraints and variables to accomodate a sink-oriented flow disaggregation. In this paper we introduce a new family of dicut collection inequalities and show that they completely describe the projection of the multicommodity formulation onto the original variables. A simple subclass is seen to include a variety of known inequalities for particular models, and combinatorial separation is examined for some special cases.

Surrogate Dual Multiplier Search Procedures in Integer Programming

Search procedures for optimal Lagrange multipliers are highly developed and provide good bounds in branch and bound procedures that have led to the successful application of Lagrangean duality in integer programming. Although the surrogate dual generally provides a better objective bound, there has been little development of surrogate multiplier search procedures. This paper develops and empirically analyzes several surrogate multiplier search procedures. Results indicate that the procedures can produce possibly superior bounds in an amount of time comparable to other techniques. Our discussion also highlights the similarity of the procedures to some well known Lagrangean search techniques.

A design methodology for fractal layout organization

This paper proposes a methodology for designing job shops under the fractal layout organization that has been introduced as an alternative to the more traditional function and product organizations. We first begin with an illustration of how a fractal job shop is constituted from individual fractal cells. We then consider joint assignment of products and their processing requirements to fractal cells, the layout of workstation replicates in a fractal cell and the layout of cells with respect to each other. The main challenge in assigning flow to workstation replicates is that flow assignment is in itself a layout dependent decision problem. We confront this dilemma by proposing an iterative algorithm that updates layouts depending on flow assignments, and flow assignments based on layout. The proposed heuristic is computationally feasible as evidenced by our experience with test problems taken from the literature. We conclude by showing how the methodologies developed in this paper have helped us evaluate fractal job shop designs through specification of fractal cells, assignment of processing requirements to workstation replicates, and development of processor level layouts. This step has had the far-reaching consequence of demonstrating the viability and the validity of the fractal layout organization.

Guaranteed performance heuristics for the bottleneck travelling salesman problem

We consider constant-performance, polynomial-time, nonexact algorithms for the minimax or bottleneck version of the Travelling Salesman Problem. It is first shown that no such algorithm can exist for problems with arbitrary costs unless P = NP. However, when costs are positive and satisfy the triangle inequality, we use results pertaining to the squares of biconnected graphs to produce a polynomial-time algorithm with worst-case bound 2 and show further that, unless P = NP, no polynomial alternative can improve on this value.