Abstract Recent work for time-series forecasting increasingly leverages the high predictive power of Deep Learning models. With this increase in model complexity, however, comes a lack in understanding of the underlying model decision process, which is problematic for safety-critical application scenarios. At the same time, simple, interpretable forecasting methods such as ARIMA and ETS still perform very well, sometimes on-par with Deep Learning approaches. We argue that using interpretable forecasters leads to good predictions in most cases. However, the forecasting performance can be improved by selecting a Deep Learning method only for few, important predictions, increasing the overall interpretability of the forecasting process. In this context, we propose a novel online model selection framework which learns to identify these predictions. An extensive empirical study on various real-world datasets containing over 3500 individual time-series shows that our selection methodology performs comparable to state-of-the-art online model selection methods in most cases while being significantly more interpretable. We find that almost always choosing a simple autoregressive or exponential smoothing model for forecasting, results in competitive performance, suggesting that the need for opaque black-box models in time-series forecasting might be smaller than recent works would suggest.

Most graph neural networks (GNNs) utilize approximations of the general graph convolution derived in the graph Fourier domain. While GNNs are typically applied in the multi-input multi-output (MIMO) case, the approximations are performed in the single-input single-output (SISO) case. In this work, we first derive the MIMO graph convolution through the convolution theorem and approximate it directly in the MIMO case. We find the key MIMO-specific property of the graph convolution to be operating on multiple computational graphs, or equivalently, applying distinct feature transformations for each pair of nodes. As a localized approximation, we introduce localized MIMO graph convolutions (LMGCs), which generalize many linear message-passing neural networks. For almost every choice of edge weights, we prove that LMGCs with a single computational graph are injective on multisets, and the resulting representations are linearly independent when more than one computational graph is used. Our experimental results confirm that an LMGC can combine the benefits of various methods.

Navigating unsignalized intersections in urban environments poses a complex challenge for self-driving vehicles, where issues such as view obstructions, unpredictable pedestrian crossings, and diverse traffic participants demand a great focus on crash prevention. In this paper, we propose a novel state representation for Reinforcement Learning (RL) agents centered around the information perceivable by an autonomous agent, enabling the safe navigation of previously uncharted road maps.Our approach surpasses several baseline models by a significant margin in terms of safety and energy consumption metrics. These improvements are achieved while maintaining a competitive average travel speed. Our findings pave the way for more robust and reliable autonomous navigation strategies, promising safer and more efficient urban traffic environments.

The lack of trust in algorithms is usually an issue when using Reinforcement Learning (RL) agents for control in real-world domains such as production plants, autonomous vehicles, or traffic-related infrastructure, partly due to the lack of verifiability of the model itself. In such scenarios, Petri nets (PNs) are often available for flowcharts or process steps, as they are versatile and standardized. In order to facilitate integration of RL models and as a step towards increasing AI trustworthiness, we propose an approach that uses PNs with three main advantages over typical RL approaches: Firstly, the agent can now easily be modeled with a combined state including both external environmental observations and agent-specific state information from a given PN. Secondly, we can enforce constraints for state-dependent actions through the inherent PN model. And lastly, we can increase trustworthiness by verifying PN properties through techniques such as model checking. We test our approach on a typical four-way intersection traffic light control setting and present our results, beating cycle-based baselines.

Abstract With machine learning (ML) becoming a popular tool across all domains, practitioners are in dire need of comprehensive reporting on the state-of-the-art. Benchmarks and open databases provide helpful insights for many tasks, however suffer from several phenomena: Firstly, they overly focus on prediction quality, which is problematic considering the demand for more sustainability in ML. Depending on the use case at hand, interested users might also face tight resource constraints and thus should be allowed to interact with reporting frameworks, in order to prioritize certain reported characteristics. Furthermore, as some practitioners might not yet be well-skilled in ML, it is important to convey information on a more abstract, comprehensible level. Usability and extendability are key for moving with the state-of-the-art and in order to be trustworthy, frameworks should explicitly address reproducibility. In this work, we analyze established reporting systems under consideration of the aforementioned issues. Afterwards, we propose STREP, our novel framework that aims at overcoming these shortcomings and paves the way towards more sustainable and trustworthy reporting. We use STREP’s (publicly available) implementation to investigate various existing report databases. Our experimental results unveil the need for making reporting more resource-aware and demonstrate our framework’s capabilities of overcoming current reporting limitations. With our work, we want to initiate a paradigm shift in reporting and help with making ML advances more considerate of sustainability and trustworthiness.

While probabilistic graphical models are a central tool for reasoning under uncertainty in AI, they are in general not as expressive as deep neural models, and inference is notoriously hard and slow. In contrast, deep probabilistic models such as sum-product networks (SPNs) capture joint distributions and ensure tractable inference, but still lack the expressive power of intractable models based on deep neural networks. In this paper, we introduce conditional SPNs (CSPNs)—conditional density estimators for multivariate and potentially hybrid domains—and develop a structure-learning approach that derives both the structure and parameters of CSPNs from data. To harness the expressive power of deep neural networks (DNNs), we also show how to realize CSPNs by conditioning the parameters of vanilla SPNs on the input using DNNs as gate functions. In contrast to SPNs whose high-level structure can not be explicitly manipulated, CSPNs can naturally be used as tractable building blocks of deep probabilistic models whose modular structure maintains high-level interpretability. In experiments, we demonstrate that CSPNs are competitive with other probabilistic models and yield superior performance on structured prediction, conditional density estimation, auto-regressive image modeling, and multilabel image classification. In particular, we show that employing CSPNs as encoders and decoders within variational autoencoders can help to relax the commonly used mean field assumption and in turn improve performance.

The exploitation of vehicles as mobile sensors acts as a catalyst for novel crowdsensing-based applications such as intelligent traffic control and distributed weather forecast. However, the massive increases in Machine-type Communication (MTC) highly stress the capacities of the network infrastructure. With the system-immanent limitation of resources in cellular networks and the resource competition between human cell users and MTC, more resource-efficient channel access methods are required in order to improve the coexistence of the different communicating entities. In this paper, we present a machine learning-enabled transmission scheme for client-side opportunistic data transmission. By considering the measured channel state as well as the predicted future channel behavior, delay-tolerant MTC is performed with respect to the anticipated resource-efficiency. The proposed mechanism is evaluated in comprehensive field evaluations in public Long Term Evolution (LTE) networks, where it is able to increase the mean data rate by 194% while simultaneously reducing the average power consumption by up to 54%.

Traffic congestion is one of the most pressing issues for smart cities. Information on traffic flow can be used to reduce congestion by predicting vehicle counts at unmonitored locations so that counter-measures can be applied before congestion appears. To do so pricy sensors must be distributed sparsely in the city and at important roads in the city center to collect road and vehicle information throughout the city in real-time. Then, Machine Learning models can be applied to predict vehicle counts at unmonitored locations. To be fault-tolerant and increase coverage of the traffic predictions to the suburbs, rural regions, or even neighboring villages, these Machine Learning models should not operate at a central traffic control room but rather be distributed across the city. Gaussian Processes (GP) work well in the context of traffic count prediction, but cannot capitalize on the vast amount of data available in an entire city. Furthermore, Gaussian Processes are a global and centralized model, which requires all measurements to be available at a central computation node. Product of Expert (PoE) models have been proposed as a scalable alternative to Gaussian Processes. A PoE model trains multiple, independent GPs on different subsets of the data and weight individual predictions based on each experts uncertainty. These methods work well, but they assume that experts are independent even though they may share data points. Furthermore, PoE models require exhaustive communication bandwidth between the individual experts to form the final prediction. In this paper we propose a hierarchical Product of Expert model, which consist of multiple layers of small, independent and local GP experts. We view Gaussian Process induction as regularized optimization procedure and utilize this view to derive an efficient algorithm which selects independent regions of the data. Then, we train local expert models on these regions, so that each expert is responsible for a given region. The resulting algorithm scales well for large amounts of data and outperforms flat PoE models in terms of communication cost, model size and predictive performance. Last, we discuss how to deploy these local expert models onto small devices.

Upcoming Intelligent Traffic Control Systems (ITSCs) will base their optimization processes on crowdsensing data obtained for cars that are used as mobile sensor nodes. In conclusion, public cellular networks will be confronted with massive increases in Machine-Type Communication (MTC) and will require efficient communication schemes to minimize the interference of Internet of Things (IoT) data traffic with human communication. In this demonstration, we present an Open Source framework for context-aware transmission of vehicular sensor data that exploits knowledge about the characteristics of the transmission channel for leveraging connectivity hotspots, where data transmissions can be performed with a high grade if resource efficiency. At the conference, we will present the measurement application for acquisition and live-visualization of the required network quality indicators and show how the transmission scheme performs in real-world vehicular scenarios based on measurement data obtained from field experiments.

In recent years we have witnessed a growing interest in trip planning systems aiming at organizing daily travel schedules in smart cities. Such systems use specialized engines to find optimal means of transport between two geospatial endpoints to provide recommendations to citizens for short routes across the city. At the same time, alternative means of transportation, such as bike sharing systems, have enjoyed tremendous success since they offer a green and facile solution for daily commuters and tourists. However, one major challenge of the bike sharing systems is that the distribution of bikes among the stations can be quite uneven during rush hours or due to topography. This often results in shortage of bikes and increasing numbers of disappointed users. Existing works in the literature are limited since they only focus on predicting the demand or apply a-posteriori methods for balancing the load of stations. Furthermore, none of these works consider the benefit of these systems in concert. In this work, we present "MOToR" (MultimOdal Trip Rebalancing), a system that builds upon the OpenTripPlanner framework to incorporate dynamic transit schedule data while balancing the availability of bikes among the bike stations. Our experimental evaluation shows that our approach is practical, efficient and outperforms state-of-the-art methods for route planning.

While cars were only considered as means of personal transportation for a long time, they are currently transcending to mobile sensor nodes that gather highly up-to-date information for crowdsensing-enabled big data services in a smart city context. Consequently, upcoming 5G communication networks will be confronted with massive increases in Machine-type Communication (MTC) and require resource-efficient transmission methods in order to optimize the overall system performance and provide interference-free coexistence with human data traffic that is using the same public cellular network. In this paper, we bring together mobility prediction and machine learning based channel quality estimation in order to improve the resource-efficiency of car-to-cloud data transfer by scheduling the transmission time of the sensor data with respect to the anticipated behavior of the communication context. In a comprehensive field evaluation campaign, we evaluate the proposed context-predictive approach in a public cellular network scenario where it is able to increase the average data rate by up to 194% while simultaneously reducing the mean uplink power consumption by up to 54%.

Upcoming 5G-based communication networks will be confronted with huge increases in the amount of transmitted sensor data related to massive deployments of static and mobile Internet of Things (IoT) systems. Cars acting as mobile sensors will become important data sources for cloud-based applications like predictive maintenance and dynamic traffic forecast. Due to the limitation of available communication resources, it is expected that the grows in Machine-Type Communication (MTC) will cause severe interference with Human-to-human (H2H) communication. Consequently, more efficient transmission methods are highly required. In this paper, we present a probabilistic scheme for efficient transmission of vehicular sensor data which leverages favorable channel conditions and avoids transmissions when they are expected to be highly resource-consuming. Multiple variants of the proposed scheme are evaluated in comprehensive realworld experiments. Through machine learning based combination of multiple context metrics, the proposed scheme is able to achieve up to 164% higher average data rate values for sensor applications with soft deadline requirements compared to regular periodic transmission.

Urban areas are increasingly subject to congestions. Most navigation systems and algorithms that avoid these congestions consider drivers independently and can, thus, cause novel congestions at unexpected places. Pre-computation of optimal trips (Nash equilibrium) could be a solution to the problem but is due to its static nature of no practical relevance. In contrast, the paper at-hand provides an approach to avoid traffic jams with dynamic self-organizing trip planning. We apply reinforcement learning to learn dynamic weights for routing from the decisions and feedback logs of the vehicles. In order to compare our routing regime against others, we validate our approach in an open simulation environment (LuST) that allows reproduction of the traffic in Luxembourg for a particular day. Additionally, in two realistic scenarios: (1) usage of stationary sensors and (2) deployment in a mobile navigation system, we perform experiments with varying penetration rates. All our experiments reveal that performance of the traffic network is increased and occurrence of traffic jams are reduced by application of our routing regime.

We present a novel approach for event recognition in massive streams of heterogeneous data driven by privacy policies and big data event processing. New technologies in mobile computing combined with sensing infrastructures distributed in a city or country are generating massive, poly-structured spatio-temporal data. With a view on emergencies and disasters these various data sources enable early response and offer situative insights when integrated in an on-line incident recognition system. Our hereby presented system architecture integrates multi-faceted sensing and distributed event detection to identify, label and increase confidence in detected incidents. A higher flexibility than existing event detection approaches is achieved by combination of the data streams at a round table. At the round table the data flow adjusts itself during execution of the real-time detection system. This offers more robustness in case streams appear or disappear. The developed architecture is used in nation-wide and city-level incident recognition scenarios.

Urban traffic gathers increasing interest as cities become bigger, crowded and “smart”. We present a system for het-erogeneous stream processing and crowdsourcing supporting intelligent urban traffic management. Complex events related to traffic congestion (trends) are detected from heterogeneous sources involving fixed sensors mounted on intersections and mobile sensors mounted on public transport vehicles. To deal with data veracity, a crowdsourcing component handles and resolves sensor disagreement. Furthermore, to deal with data sparsity, a traffic modelling component offers information in areas with low sensor coverage. We demonstrate the system with a real-world use-case from Dublin city, Ireland.

Recently evolved Bluetooth tracking technology is currently applied to extract individual pathways, movement patterns or to rank popularity of locations by their visitor quantities. To utilize this technology for the creation of location aware intelligent environments, the next steps are to come up with microscopic traffic values. This work proposes a solution for this question, namely, a Bluetooth localization and path reproduction of individual persons. Our approach applies radio signal strength fingerprinting. We introduce and evaluate this approach for varying granularities and mobile phone types. The result is an accurate reproduction of pedestrian position and route choice in a complex facility.