Abstract
An unsupervised method for convolutional neural network (CNN) architecture design is proposed. The method relies on a variable neighborhood search-based approach for finding CNN architectures and hyperparameter values that improve classification performance. For this purpose, t-Distributed Stochastic Neighbor Embedding (t-SNE) is applied to effectively represent the solution space in 2D. Then, k-Means clustering divides this representation space having in account the relative distance between neighbors. The algorithm is tested in the CIFAR-10 image dataset. The obtained solution improves the CNN validation loss by over 15% and the respective accuracy by 5%. Moreover, the network shows higher predictive power and robustness, validating our method for the optimization of CNN design.

Abstract
Allocation of jobs to machines and subsequent sequencing each machine is known as job scheduling problem. Classically, both operations are done in a centralized and static/offline structure, considering some assumptions about the jobs and machining environment. Today, with the advent of Industry 4.0, the need to incorporate real-time data in the scheduling decision process is clear and facilitated. Recently, several studies have been conducted on the collection and application of distributed data in real-time of operations, e.g., job scheduling and control. In practice, pure distribution and decentralization is not yet fully realizable because of e.g., transformation complexity and classical resistance to change. This paper studies a combination of decentralized sequencing and central optimum allocation in a lithography job-shop problem. It compares the level of applicability of two decentralized algorithms against the central scheduling. The results show better relative performance of sequencing in stochastic cases. © IFIP International Federation for Information Processing 2017.

Abstract
The pharmaceutical industry operates in a very competitive and regulated market The increased pressure of pharmacies to order fewer products and to receive them more frequently is overcharging the pharmaceutical's distribution network Furthermore, the tight margins and the continuous growth of generic drugs consumption are pressing wholesalers to optimize their supply chains. In order to survive, wholesalers are rethinking their strategies to increase competitiveness. This paper proposes an optimization-simulation approach to address the wholesalers network redesign problem, trading off the operational costs and customer service level. Firstly, at a strategic-tactical level, the supply chain network redesign decisions are optimized via a mixed integer programming model. Here, the number, location, function and capacity of the warehouses, the allocation of customers to the warehouses and the capacity and function of the distribution channels are defined. Secondly, at an operation level, the solution found is evaluated by means of a discrete event simulation model to assess the impact of the redesign in the wholesaler's daily activities. Computational results on a pharmaceutical wholesaler case-study are discussed and the benefits of this solution approach exposed.

Abstract
Tactical production-distribution "planning models have attracted a great deal of attention in the past decades. In these models, production and distribution decisions are considered simultaneously such that the combined plans are more advantageous than the plans resolved in a hierarchical planning process. We consider a two-stage production process, where in the first stage raw materials are transformed into continuous resources that feed the discrete production of end products in the second stage. Moreover, the setup times and costs of resources depend on the sequence in which they are processed in the first stage. The minimum scheduling unit is the product family which consists of products sharing common resources and manufacturing processes. Based on different mathematical modelling approaches to the production in the first stage, we develop a sequence-oriented formulation and a product-oriented formulation, and propose decomposition-based heuristics to solve this problem efficiently. By considering these dependencies arising in practical production processes, our model can be applied to various industrial cases, such as the beverage industry or the steel industry. Computation tests on instances from an industrial application are provided at the end of the paper.

Abstract

Abstract
The hierarchical planning concept is commonly used for production planning. Dividing the planning process into subprocesses which are solved separately in the order of the hierarchy decreases the complexity and fits the common organizational structure. However, interaction between planning levels is crucial to avoid infeasibility and inconsistency of plans. Furthermore, optimizing subproblems often leads to suboptimal results for the overall problem. The alternative, a monolithic model integrating all planning levels, has been rejected in the literature because of several reasons. In this study, we show that some of them do not hold for an integrated production planning model combining the planning tasks usually attributed to aggregate production planning and master production scheduling. Therefore, we develop a hierarchical and an integrated model considering both levels, aggregate production planning and master production scheduling. Computational tests show that it is possible to solve the integrated model and that it outperforms the hierarchical approach for all instances. Moreover, an indication is given why and when integration is beneficial.