COAP 2015 Best Paper Prize

Each year, the editorial board of computational optimization and applications (COAP) selects a paper from the preceding year’s publications for the Best Paper Award. In 2015, 92 papers were published by COAP. The recipients of the 2015 Best Paper Award are Qi Huangfu of the FICO Xpress Team and Julian Hall of the University of Edinburgh for their paper “Novel update techniques for the revised simplex method” published in volume 60, pages 587–608. We point out that Julian Hall has now won 3 of the 14 Best Paper Awards given out since the inception of the award in 2003, a remarkable achievement. This article highlights the research related to the award winning paper. The paper introduces three novel techniques for updating the invertible representation of the basis matrix when solving practical sparse linear programming (LP) problems using a high performance implementation of the dual revised simplex method. These are of particular value when suboptimization is used. Two are variants of the product form update and the other permits multiple Forrest-Tomlin updates to be performed. The development of an efficient parallel implementation of the revised simplex method has attracted the attention of leading researchers over the past 30 years, butwith relatively little success [4] until recently. For example, Hall and McKinnon devised two parallel schemes for the primal revised simplex method in the mid 1990s [6,7]. Their moderately efficient prototype implementation gave limited speedup on general sparse LP problems, but the performance was well below that of the best serial solvers of the time. A byproduct of this work was the identification of hyper-sparsity in the simplex method and techniques for its exploitation using serial computing [8], work which received the COAP Best Paper Award for 2005. A few years ago, Lubin, Hall, Petra and Anitecu developed a parallelization of the revised simplex method for large scale two-stage stochastic LP problems [10]. The performance of their distributed memory application scaled linearly on hundreds of nodes and enabled LP problems with hundreds of millions of variables to be solved and they received the COAP Best Paper Award for 2013. However, for general large scale sparse LP problems, the search