Fakultät Informatik und Mathematik
Refine
Document Type
- conference proceeding (article) (11) (remove)
Language
- English (11)
Has Fulltext
- yes (11) (remove)
Is part of the Bibliography
- no (11)
Keywords
- Adam optimizer (2)
- Convergence (2)
- Konvergenz 〈Informationstechnik〉 (2)
- Maschinelles Lernen (2)
- Neuronales Netz (2)
- Optimierungsalgorithmus (2)
- Auftragsabwicklung (1)
- Clearing function (1)
- Cloud Computing (1)
- Computer algebra (1)
Institute
Begutachtungsstatus
- peer-reviewed (11) (remove)
In this paper a new method for placing bus stops is presented. The method is suitable for permanently installed new bus stops and temporarily chosen collection points for call busses as well. Moreover, our implementation of the Voronoi algorithm chooses new locations for bus stops in such a way that more bus stops are set in densely populated areas and less in less populated areas. To achieve this goal, a corresponding weighting is applied to each possible placement point, based on the number of inhabitants around this point and the points of interest, such as medical centers and department stores around this point. Using the area of Roding, a small town in Bavaria, for a case study, we show that our method is especially suitable for for rural areas, where there are few multi-family houses or apartment blocks and the area is not densely populated.
In the realm of parallel computing, optimization plays a pivotal role in achieving efficient and scalable solutions. In this work, we present the parallelization of a hybrid genetic search for solving the Capacitated Vehicle Routing Problem with Pickup and Delivery (CVRPPD).It leverages the synergy between genetic algorithms and parallel computing to address the complex optimization problem. This hybrid algorithm combines a customized version of local search with a genetic algorithm to compute an effective solution. Our implementation makes use of the Message Passing Interface (MPI) for data distribution and parallel execution. In addition, we run multi-threaded processes on NVIDIA graphical processors using the CUDA technology, which further increases the computation speed and consequently minimizes the runtime. Parallelization also allows the best-improvement strategy to be used instead of the rst-improvement strategy while maintaining the same runtime. We store the resulting routes in a bus route database which we created as the basis of an extensive library of optimal routes for our specifc use case of optimizing bus routes in a rural area. The experimental results on real road data show that the parallel implementation of the Hybrid Genetic Search (HGS) achieves significant improvements in runtime over the sequential implementation above a certain problem size. We believe that our implementation of the parallel hybrid genetic search method can have a great in influence on optimization strategies in parallel computing and can also be applied to other subproblems of the VRP.
Adaptive Moment Estimation (Adam) is a very popular training algorithm for deep neural networks, implemented in many machine learning frameworks. To the best of the authors knowledge no complete convergence analysis exists for Adam. The contribution of this paper is a method for the local convergence analysis in batch mode for a deterministic fixed training set, which gives necessary conditions for the hyperparameters of the Adam algorithm. Due to the local nature of the arguments the objective function can be non-convex but must be at least twice continuously differentiable.
In this article, we discuss energy consumption of producing firms on aggregate production planning. While almost constant energy consumption can be the case for a producing firm, highly fluctuating energy demand can occur as well. Together with volatile energy supply, e.g. due to renewable energy sources, this combination of fluctuating energy supply and demand can result in planning uncertainty and high energy costs. We propose different case studies in which such high deviation in the electricity consumption of a producing firm occurs due to aggregate production planning without appropriate consideration of energy consumption.
Flexibility and in particular volume flexibility is an important topic for industrial manufacturing companies. In this context, the harmonization of the available and required capacity is a central task, especially with increasing fluctuations in customer demand. In classical approaches , this is considered only by the use of additional capacities and there are only a few approaches that combine aspects of personnel planning with production planning. Therefore, this article presents a linear optimization model for master production scheduling that includes aspects of personnel requirements planning. It is used to investigate different strategies for the use of overtime and temporary workers in order to achieve different levels of volume flexibility. With regard to the monetary and social impacts, the results indicate that overtime has a stronger influence to achieve volume flexibility than the use of temporary workers. However, both are affected by substantial deficits in human working conditions. But the results also imply a promising potential for improving the social aspects without a significant increase in costs.
Most large-scale organizations adopted Cloud Computing (CC) on a company level in recent years. Managers now face the challenge to appropriately implement CC "operationally", i.e., for information systems (ISs). We refer to this as post-adoption, addressing the extent of technology usage after adoption. Specifically, managers need to choose among the CC delivery models Infrastructure-as-a-Service (IaaS), Platform-as-a-Service (PaaS), and Software-asa-Service (SaaS). We differentiate the determinants of this post-adoption decision for IaaS, PaaS, and SaaS. Based on this analysis, we derive criteria that guide managers' delivery model selection: Adopt 1) IaaS for ISs requiring flexibility and reduced time to market, 2) PaaS to access specialized resources, and 3) SaaS to focus on core competencies. Moreover, we analyze the impact on the CC strategy and postulate them as recommendations: I) acknowledge the interplay between governance and time-to-market, II) realize cost savings on company level, and III) consider strategically important ISs for CC.
Sustainability is an important topic in production plan-ning and control. This article contributes in particular the to further research on the social dimension. It pre-sents a linear optimisation model for Master Produc-tion Scheduling in order to improve human working conditions. Existing approaches have already identified a considerable potential for improvements. Further-more, this article analyses the influence of the compa-ny size on workload and costs using an application with a high proportion of manual activities. It is demonstrat-ed that human working conditions can be improved independently from the company size without increas-ing costs. In addition, smaller companies tend to have a higher exhaustion and the workload affects the total costs more in smaller companies. Therefore, smaller companies might benefit more from an improvement in human working conditions.
This paper connects research from business model innovation and information systems by exploring critical IT capabilities for servitized business models. The adoption of servitized business models is a major business model innovation strategy. At the same time, digitalization drives the evolution of IT capabilities at these business models. Scholars argue that it remains unclear how IT capabilities enable servitized business models to build a competitive advantage by achieving cost advantages or differentiation. This paper explores IT capabilities that enable building a competitive advantage for servitized business models based on a qualitative analysis of multiple published case studies. The authors identify configurations of IT capabilities among servitized business models. The findings contribute to servitization research by exploring IT capabilities
and how they are combined among servitized business models.
The insights help practitioners deploy digital technologies and IT
assets effectively as building blocks of IT capabilities to advance
their servitized business model.
Identifying different functional regions during a brain surgery is a challenging task usually performed by highly specialized neurophysiologists. Progress in this field may be used to improve in situ brain navigation and will serve as an important building block to minimize the number of animals in preclinical brain research required by properly positioning implants intraoperatively. The study at hand aims to correlate recorded extracellular signals with the volume of origin by deep learning methods. Our work establishes connections between the position in the brain and recorded high-density neural signals. This was achieved by evaluating the performance of BLSTM, BGRU, QRNN and CNN neural network architectures on multisite electrophysiological data sets. All networks were able to successfully distinguish cortical and thalamic brain regions according to their respective neural signals. The BGRU provides the best results with an accuracy of 88.6 % and demonstrates that this classification task might be solved in higher detail while minimizing complex preprocessing steps.
One of the most popular training algorithms for deep neural networks is the Adaptive Moment Estimation (Adam) introduced by Kingma and Ba. Despite its success in many applications there is no satisfactory convergence analysis: only local convergence can be shown for batch mode under some restrictions on the hyperparameters, counterexamples exist for incremental mode. Recent results show that for simple quadratic objective functions limit cycles of period 2 exist in batch mode, but only for atypical hyperparameters, and only for the algorithm without bias correction. We extend the convergence analysis to all choices of the hyperparameters for quadratic functions. This finally answers the question of convergence for Adam in batch mode to the negative. We analyze the stability of these limit cycles and relate our analysis to other results where approximate convergence was shown, but under the additional assumption of bounded gradients which does not apply to quadratic functions. The investigation heavily relies on the use of computer algebra due to the complexity of the equations.