In many cases of machine learning, research suggests that the development of training data might have a higher relevance than the choice and modelling of classifiers themselves. Thus, data augmentation methods have been developed to improve classifiers by artificially created training data. In NLP, there is the challenge of establishing universal rules for text transformations which provide new linguistic patterns. In this paper, we present and evaluate a text generation method suitable to increase the performance of classifiers for long and short texts. We achieved promising improvements when evaluating short as well as long text tasks with the enhancement by our text generation method. Especially with regard to small data analytics, additive accuracy gains of up to 15.53% and 3.56% are achieved within a constructed low data regime, compared to the no augmentation baseline and another data augmentation technique. As the current track of these constructed regimes is not universally applicable, we also show major improvements in several real world low data tasks (up to +4.84 F1-score). Since we are evaluating the method from many perspectives (in total 11 datasets), we also observe situations where the method might not be suitable. We discuss implications and patterns for the successful application of our approach on different types of datasets.
Traffic routing is a well-established optimization problem in traffic management. Here, we address dynamic routing problems where the load of roads is taken into account dynamically, aiming at the optimization of required travel times. We investigate ant-based algorithms that can handle dynamic routing problems, but suffer from negative emergent effects like road congestions. These negative effects are inherent in the design of ant-based algorithms. In this article we propose an inverse ant-based routing algorithm to (a) maintain the positive features of ant-based algorithms for dynamic routing problems, while (b) avoiding the occurrence of negative emerging effects, like road congestion. We evaluated the performance of the proposed algorithm by comparing its results with two alternative routing algorithms, namely, A * , which is a static routing algorithm, and an iterative approach. In particular, the iterative approach is used for providing an upper bound, as it uses routing knowledge in a number of calibration runs, to determine the actual load, before the effective routing is done. For the evaluation we used the agent-based traffic simulation system MAINSIM. The evaluation was done with one synthetic and two real-world scenarios, to outline the practical relevance of our findings. Based on these evaluations, we can conclude that the inverse ant-based routing approach is particularly suited for a scenario with a high traffic density, as it can adapt the routing of each vehicle, while avoiding the negative emerging effects of conventional ant-based routing algorithms.
Simulations are widely used for modeling, analysis, planning, and optimisation of traffic flows and phenomena. For realistic traffic simulations within urban scenarios, the following tasks have to be solved: (1) modeling of the road structure; (2) specification of the behaviour on the road. In our days, very detailed road models for almost any major city exist in Geographic Information Systems (GIS). In the last two decades, the Nagel-Schreckenberg model (NaSch) has been established as de facto standard for car behaviour in freeway traffic due to its efficient and realistic simulations. Within urban scenarios, NaSch lacks of flexibility to integrate heterogeneous road users like cars and bicycles. The tasks mentioned before are addressed in this paper, i.e., we propose an approach for modeling and specification of urban mixed traffic simulations. As a first step (1), an extended graph as basis for traffic simulation has to be designed. For a concrete scenario, it will be automatically generated on basis of OpenStreetmap cartographical material. The specification of road user behaviour (2) has been influenced by the NaSch model. However, the model has been extended to cover the lack of NaSch in urban scenarios: A non cell-based approach is chosen for traffic movement. Furthermore, the routing of traffic users is based on either probability or A* based routing. In this paper, details on the modeling and specification are presented and experimental results are provided.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.