diff --git a/content/news/2025/seminars/january_to_july/bargetto.md b/content/news/2025/seminars/january_to_july/bargetto.md index 9d9ca298995dabc2e8017d3552a9b470fc8617b8..4277be53fed30c6c5df73bda2f9147705c5b7cb1 100644 --- a/content/news/2025/seminars/january_to_july/bargetto.md +++ b/content/news/2025/seminars/january_to_july/bargetto.md @@ -1,5 +1,5 @@ --- -title: To be announced +title: "Iterated Inside Out and Its Specialized Variant for DOTmark Instances: A New Exact Algorithm for the Transportation Problem" speaker: Roberto Bargetto tags: [seminar, seminars_2025_jan_july] date: 2025-04-03 # very important @@ -15,5 +15,11 @@ image: imagesize: 30 # do not change --- - -To be announced +The transportation problem has recently gained renewed interest due to its applications in machine learning and artificial intelligence, particularly in measuring distances between images. +To address the associated computational challenges, we propose a novel algorithm, Iterated Inside Out, for the exact resolution of the well-known transportation problem. +The core of our algorithm requires an initial basic feasible solution and consists of two iteratively repeated phases until an optimal basic feasible solution is obtained. +In the "inside" phase, the algorithm progressively improves the current solution by increasing the value of non-basic variables with negative reduced costs, typically yielding a feasible solution that is non-basic and interior to the constraint set polytope. +In the "outside" phase, the algorithm improves the current solution by iteratively setting variables to zero until a new improved basic feasible solution is reached. +Building on this core algorithm, we develop a specialized variant tailored for image processing. +Extensive computational experiments demonstrate that our approach significantly outperforms commercial solvers such as CPLEX and Gurobi, as well as other exact algorithms in the literature. +We demonstrate its superiority on both randomly generated instances and the DOTmark benchmark, a standard dataset for image processing and large-scale transportation problems.