EURO Journal on Computational Optimization最新文献

Solving large-scale Mixed Integer Linear Programs (MIP) can be difficult without advanced algorithms such as decomposition based techniques. Even if a decomposition technique might be appropriate, there are still many possible decompositions for any large MIP and it may not be obvious which will be the most effective. The quality of a decomposition depends on both the tightness of the dual bound, in our case generated via Lagrangian Relaxation, and the computational time required to produce that bound. Both of these factors are difficult to predict, motivating the use of a Machine Learning function to predict decomposition quality based a score that combines both bound quality and computational time. This paper presents a comprehensive analysis of the predictive capabilities of a ML function for predicting the quality of MIP decompositions created via constraint relaxation. In this analysis, the role of instance similarity and ML prediction quality is explored, as well as the benchmarking of a ML ranking function against existing heuristic functions. For this analysis, a new dataset consisting of over 40000 unique decompositions sampled from across 24 instances from the MIPLIB2017 library has been established. These decompostions have been created by both a greedy relaxation algorithm as well as a population based multi-objective algorithm, which has previously been shown to produce high quality decompositions. In this paper, we demonstrate that a ML ranking function is able to provide state-of-the-art predictions when benchmarked against existing heuristic ranking functions. Additionally, we demonstrate that by only considering a small set of features related to the relaxed constraints in each decomposition, a ML ranking function is still able to be competitive with heuristic techniques. Such a finding is promising for future constraint relaxation approaches, as these features can be used to guide decomposition creation. Finally, we highlight where a ML ranking function would be beneficial in a decomposition creation framework.

We present and compare novel binary programs for linear ordering problems that involve the notion of asymmetric betweenness and expose relations to the quadratic linear ordering problem and its linearization. While two of the binary programs prove particularly superior from a computational point of view when many or all betweenness relations shall be modeled, the others arise as natural formulations that resemble important theoretical correspondences and provide a compact alternative for sparse problem instances. A reasoning for the strengths and weaknesses of the different formulations is derived by means of polyhedral considerations with respect to their continuous relaxations.

In this article we propose a binary classification model to distinguish a specific class that corresponds to a characteristic that we intend to identify (fraud, spam, disease). The classification model is based on a cloud of spheres that circumscribes the points of the class to be identified. It is intended to build a model based on a cloud and not on a disjoint set of clouds, establishing this condition on the connectivity of a graph induced by the spheres. To solve the problem, designed by a Cloud of Connected Spheres, a quadratic model with continuous and binary variables (MINLP) is proposed with the minimization of the number of spheres. The issue of connectivity implies in many models the imposition of an exponential number of constraints. However, because of the specific conditions of the problem under study, connectivity is enforced with linear constraints that scale quadratically with K, which serves as an upper bound on the number of spheres. This classification model is effective when the structure of the class to be identified is highly non-linear and non-convex, also adapting to the case of linear separation. Unlike neural networks, the classification model is transparent, with the structure perfectly identified. No kernel functions are used and it is not necessary to use meta-parameters unless it is intended also to maximize the separation margin as it is done in SVM. Finding the global optima for large instances is quite challenging, and to address this, a heuristic is proposed. The heuristic demonstrates nice results on a set of frequently tested real problems when compared to state-of-the-art algorithms.

The Feasibility Pump (FP) is one of the best-known primal heuristics for mixed-integer programming (MIP): more than 15 papers suggested various modifications of all of its steps. So far, no variant considered information across multiple iterations, but all instead maintained the principle to optimize towards a single reference integer point. In this paper, we evaluate the usage of multiple reference vectors in all stages of the FP algorithm. In particular, we use LP-feasible vectors obtained during the main loop to tighten the variable domains before entering the computationally expensive enumeration stage, a procedure we refer to as mRENS. Moreover, we consider multiple integer reference vectors to explore further optimizing directions and introduce alternative objective scaling terms to balance the contributions of the distance functions and the original MIP objective.

Our computational experiments demonstrate that the new method can improve performance on general MIP test sets. In detail, our modifications provide a 29.3% solution quality improvement and 4.0% running time improvement in an embedded setting, needing 16.0% fewer iterations over a large test set of MIP instances. In addition, the method's success rate increases considerably within the first few iterations. In a standalone setting, we also observe a moderate performance improvement, which makes our version of FP suitable for the two main use-cases of the algorithm.

The Submodular Bin Packing (SMBP) problem asks for packing unsplittable items into a minimal number of bins for which the capacity utilization function is submodular. SMBP is equivalent to chance-constrained and robust bin packing problems under various conditions. SMBP is a hard binary nonlinear programming optimization problem. In this paper, we propose a branch-and-price algorithm to solve this problem. The resulting price subproblems are submodular knapsack problems, and we propose a tailored exact branch-and-cut algorithm based on a piece-wise linear relaxation to solve them. To speed up column generation, we develop a hybrid pricing strategy to replace the exact pricing algorithm with a fast pricing heuristic. We test our algorithms on instances generated as suggested in the literature. The computational results show the efficiency of our branch-and-price algorithm and the proposed pricing techniques.

Drones are currently seen as a viable way for improving the distribution of parcels in urban and rural environments, while working in coordination with traditional vehicles like trucks. In this paper we consider the parallel drone scheduling vehicle routing problem, where the service of a set of customers requiring a delivery is split between a fleet of trucks and a fleet of drones. We consider two variations of the problem. In the first one, the problem is more theoretical, and the target is the minimization of the time required to complete the service and have all the vehicles back to the depot. In the second variant, more realistic constraints involving operating costs, capacity limitation and workload balance, are considered, and the target is to minimize the total operational costs. We propose different constraint programming models to deal with the two problems. An experimental champaign on the instances previously adopted in the literature is presented to validate the new solving methods. The results show that, on top of being a viable way to solve problems to optimality, the models can also be used to derive effective heuristic solutions and high-quality lower bounds for the optimal cost, if the execution is interrupted before its natural end.

Optimization of storage using neural networks is now commonly achieved by solving a single optimization problem. We first show that this approach allows solving high-dimensional storage problems, but is limited by memory issues. We propose a modification of this algorithm based on the dynamic programming principle and propose neural networks that outperform classical feedforward networks to approximate the Bellman values of the problem. Finally, we study the stochastic linear case and show that Bellman values in storage problems can be accurately approximated using conditional cuts computed by a very recent neural network proposed by the author. This new approximation method combines linear problem solving by a linear programming solver with a neural network approximation of the Bellman values.

In this paper we deal with a problem associated with frequency assignment. Suppose we have a number of transmitters, each of which has been allocated a frequency. The problem we consider is how, given one (or more) transmitters are requesting a new frequency allocation, for example because of the interference they are currently suffering, to decide the new frequencies. Here we wish to constrain overall interference, but minimise the number of frequency changes needed for transmitters that have not requested a change.

We present an optimisation model for frequency allocation that minimises changes in the existing allocation, whilst limiting interference. We consider the standard mathematical representation of interference in the literature and show that we can represent it in a way that involves far fewer variables and constraints.

We make use of this new representation of interference in our zero-one integer linear program for deciding a new frequency allocation. We also show how our formulation can be adapted to deal with a number of other possibilities, specifically allocating frequencies to new transmitters with known locations and also deciding a location (and frequency) for a single new transmitter.

We present computational results for our approach making use of minimum interference frequency assignment test problems taken from the literature. We compare the results from our new representation of interference with those obtained using the standard representation.

Designing a supply chain to comply with environmental policy requires awareness of how work and/or production methods impact the environment and what needs to be done to reduce those environmental impacts and make the company more sustainable. This is a dynamic process that occurs at both the strategic and operational levels. However, being environmentally friendly does not necessarily mean improving the efficiency of the system at the same time. Therefore, when allocating a production budget in a supply chain that implements the green paradigm, it is necessary to figure out how to properly recover costs in order to improve both sustainability and routine operations, offsetting the negative environmental impact of logistics and production without compromising the efficiency of the processes to be executed. In this paper, we study the latter problem in detail, focusing on the CO₂ emissions generated by the transportation from suppliers to production sites, and by the production activities carried out in each plant. We do this using a novel mathematical model that has a quadratic objective function and all linear constraints except one, which is also quadratic, and models the constraint on the budget that can be used for green investments caused by the increasing internal complexity created by large production flows in the production nodes of the supply network. To solve this model, we propose a multistart algorithm based on successive linear approximations. Computational results show the effectiveness of our proposal.

In this paper, we study the behavior of feasible rounding approaches for mixed-integer optimization problems when integrated into branch-and-bound methods. Our research addresses two important aspects. First, we develop insights into how an (enlarged) inner parallel set, which is the main component for feasible rounding approaches, behaves when we move down a search tree. Our theoretical results show that the number of feasible points obtainable from the inner parallel set is nondecreasing with increasing depth of the search tree. Thus, they hint at the potential benefit of integrating feasible rounding approaches into branch-and-bound methods. Second, based on those insights, we develop a novel primal heuristic for MILPs that fixes variables in a way that promotes large inner parallel sets of child nodes.

Our computational study shows that combining feasible rounding approaches with the presented diving ideas yields a significant improvement over their application in the root node. Moreover, the proposed method is able to deliver best solutions for the MIP solver SCIP for a significant share of problems which hints at its potential to support solving MILPs.