While machine learning is traditionally a resource intensive task, embedded systems, autonomous navigation, and the vision of the Internet of Things fuel the interest in resource-efficient approaches. These approaches aim for a carefully chosen trade-off between performance and resource consumption in terms of computation and energy. The development of such approaches is among the major challenges in current machine learning research and key to ensure a smooth transition of machine learning technology from a scientific environment with virtually unlimited computing resources into everyday's applications. In this article, we provide an overview of the current state of the art of machine learning techniques facilitating these real-world requirements. In particular, we focus on resource-efficient inference based on deep neural networks (DNNs), the predominant machine learning models of the past decade. We give a comprehensive overview of the vast literature that can be mainly split into three non-mutually exclusive categories: (i) quantized neural networks, (ii) network pruning, and (iii) structural efficiency. These techniques can be applied during training or as post-processing, and they are widely used to reduce the computational demands in terms of memory footprint, inference speed, and energy efficiency. We also briefly discuss different concepts of embedded hardware for DNNs and their compatibility with machine learning techniques as well as potential for energy and latency reduction. We substantiate our discussion with experiments on well-known benchmark data sets using compression techniques (quantization, pruning) for a set of resource-constrained embedded systems, such as CPUs, GPUs and FPGAs. The obtained results highlight the difficulty of finding good trade-offs between resource efficiency and prediction quality.

This paper presents a framework which allows urban planners to navigate and interact with large datasets fused with social feeds in real-time, enhanced by a virtual reality (VR) capability, which further promotes the knowledge discovery process and allows to interact with urban data in natural yet immersive way. A challenge in urban planning is making decisions based on datasets which are many times ambiguous, together with effective use of newly available yet unstructured sources of information like social media. Providing expert users with novel ways of representing knowledge can be beneficial for decision making. Game engines have evolved into capable testbeds for novel visualization and interaction techniques. We therefore explore the possibility of using a modern game engine as a platform for knowledge representation in urban planning and how it can be used to model ambiguity. We also investigate how urban planners can benefit from immersion when it comes to data exploration and knowledge discovery. We apply the concept of using primitives to publicly available transportation datasets and social feeds of New York city, we discuss a gesture-based VR extension of our framework and lastly, we conclude the paper with feedback from expert users in urban planning and with an outlook of future challenges.

Many contemporary research tasks require complex decision making and interdisciplinary collaboration among different groups of experts and stakeholders. Nowhere is this more apparent than in the planning and design of future cities - rapidly growing cities where the development and allocation of energy, housing, transportation and other scarce resources requires the active participation of architects, urban planners, government stakeholders and private citizens. In this paper, we show how recent advances in display technology can facilitate such collaboration and create new opportunities for participatory urban planning. We introduce the Value Lab Asia as a model for collaborative planning and value creation, and show how a multi-screen display management framework called the Tool Library integrates the laboratory's physical and software infrastructure with the planning process.

This recent advances in display and user interface technologies create new opportunities for participatory urban planning. Whereas touch screens enable direct screen interactions for intuitive remodeling of urban simulations, high resolution video walls offer true-to-life or human-scale scenario visualizations. Multi-touch screen environments finally function as an interactive user interface for the entire urban simulation pipeline. However intuitive and preferably natural interactions for simple reconfiguration, overarching tool management or complex tool interaction remain a major challenge. For this reason we have implemented a gesture recognition framework that uses multimodal input devices to satisfy all gesture use cases and integrates a display management system for seamless interaction with our multi-screen infrastructure. To understand the implications of this novel approach, we conducted preliminary user trials with urban planning experts and technicians. The results indicated that the system provides a fluid and natural user experience for exploring and analyzing urban planning data.

This paper introduces the concept of dynamic Multi-View, Multi-Format, Multi-User Visualizations. It also suggests aligning all visualization branches to a common problem, namely the design and implementation of future sustainable cities, such as Jakarta, Indonesia. While a lot of visualization research has extensively discussed the emotional, cognitive and coordinative benefits of visual representations, the application of such tools to solve grand societal problems has been neglected. We suggest focusing on Future Cities, since they require solutions and because the field seems to be ideal to align the various subgroups of visualization research. We derived this insight from about 1000 events in our physical value lab and our own software development - the Visual Manager and the Shuffler. The Visual Manager uses visual metaphors to illustrate risk assessment scenarios. The Shuffler is a framework to create coordinated multiple views. The innovative part of this framework is that different visual representations, complementary business logics and datasets can be distributed to different views, which calls for more transdisciplinary work in the design of dynamic multi-view, multi-format and multi-user software. This paper is relevant for researchers in all subgroups of visualization research - especially Knowledge Visualization and Information Visualization.

The MUGGES mobile service prosuming platform aims at enabling user-aware service provision and consumption from the very mobile devices in a true peer-to-peer manner, still providing centralized support to user- and location-aware micro-service searching. Thus, users will not only provide contents but also launch and host services from their own mobile devices. Explorative field trials in Spain and Finland assessing the potential of the ‘prosuming concept' were carried out for several weeks with students and IT professionals. The feedback gathered from end-users showed that the MUGGES proposed location concept eases service creation and consumption, although service provisioning from mobile devices is still in its infancy.