Federated learning (FL) is quickly becoming the de facto standard for the distributed training of deep recommendation models, using on-device user data and reducing server costs. In a typical FL process, a central server tasks end-users to train a shared recommendation model using their local data. The local models are trained over several rounds on the users' devices and the server combines them into a global model, which is sent to the devices for the purpose of providing recommendations. Standard FL approaches use randomly selected users for training at each round, and simply average their local models to compute the global model. The resulting federated recommendation models require significant client effort to train and many communication rounds before they converge to a satisfactory accuracy. Users are left with poor quality recommendations until the late stages of training. We present a novel technique, FedFast, to accelerate distributed learning which achieves good accuracy for all users very early in the training process. We achieve this by sampling from a diverse set of participating clients in each training round and applying an active aggregation method that propagates the updated model to the other clients. Consequently, with FedFast the users benefit from far lower communication costs and more accurate models that can be consumed anytime during the training process even at the very early stages. We demonstrate the efficacy of our approach across a variety of benchmark datasets and in comparison to state-of-the-art recommendation techniques.

Admission control plays a very important role in wireless systems, as it is one of the basic mechanisms for ensuring the quality of service offered to users. Based on the available network resources, it estimates the impact of adding or dropping a new session request. In both 2G and 3G systems, admission control refers to a single network. As we are moving towards heterogeneous wireless networks referred to as systems beyond 3G or 4G, admission control will need to deal with many heterogeneous networks and admit new sessions to a network that is most appropriate to supply the requested QoS. In this article we present the fundamentals of access-network-based admission control, an overview of the existing admission control algorithms for 2G and 3G networks, and finally give the design of a new admission control algorithm suitable for future 4G networks and specifically influenced by the objectives of the European WINNER project.

Background: The increasing prevalence of sedentary lifestyles has prompted the development of innovative public health interventions, such as smartphone apps that deliver personalized exercise programs. The widespread availability of mobile technologies (eg, smartphone apps and wearable activity trackers) provides a cost-effective, scalable way to remotely deliver personalized exercise programs to users. Using machine learning (ML), specifically reinforcement learning (RL), may enhance user engagement and effectiveness of these programs by tailoring them to individual preferences and needs. Objective: The primary aim was to investigate the impact of the Samsung-developed i80 BPM app, implementing ML for exercise prescription, on user satisfaction and exercise intensity among the general population. The secondary objective was to assess the effectiveness of ML-generated exercise programs for remote prescription of exercise to members of the public. Methods: Participants were randomized to complete 3 exercise sessions per week for 12 weeks using the i80 BPM mobile app, crossing over weekly between intervention and control conditions. The intervention condition involved individualizing exercise sessions using RL, based on user preferences such as exercise difficulty, selection, and intensity, whereas under the control condition, exercise sessions were not individualized. Exercise intensity (measured by the 10-item Borg scale) and user satisfaction (measured by the 8-item version of the Physical Activity Enjoyment Scale) were recorded after the session. Results: In total, 62 participants (27 male and 42 female participants; mean age 43, SD 13 years) completed 559 exercise sessions over 12 weeks (9 sessions per participant). Generalized estimating equations showed that participants were more likely to exercise at a higher intensity (intervention: mean intensity 5.82, 95% CI 5.59‐6.05 and control: mean intensity 5.19, 95% CI 4.97‐5.41) and report higher satisfaction (RL: mean satisfaction 4, 95% CI 3.9-4.1 and baseline: mean satisfaction 3.73, 95% CI 3.6-3.8) in the RL model condition. Conclusions: The findings suggest that RL can effectively increase both the intensity with which people exercise and their enjoyment of the sessions, highlighting the potential of ML to enhance remote exercise interventions. This study underscores the benefits of personalized exercise prescriptions in increasing adherence and satisfaction, which are crucial for the long-term effectiveness of fitness programs. Further research is warranted to explore the long-term impacts and potential scalability of RL-enhanced exercise apps in diverse populations. This study contributes to the understanding of digital health interventions in exercise science, suggesting that personalized, app-based exercise prescriptions may be more effective than traditional, nonpersonalized methods. The integration of RL into exercise apps could significantly impact public health, particularly in enhancing engagement and reducing the global burden of physical inactivity.

The IoT (Internet-of-Things) concept has been introduced as a strategic innovation aspect that will benefit society in many ways. Environmental monitoring, smart traffic control, pollution monitoring, crime prevention, smart metering, etc., have now been made feasible due to the ongoing research and development towards IoT. Smart Objects (SOs) are among the fundamental blocks of an IoT architecture consisting of a sensing element (sensor) that senses the environment, and other software/hardware entities. SOs are interconnected forming wireless sensor networks (WSNs) that often convey sensitive information in a multi-hop fashion. As SOs are severe resource-constrained devices, energy-efficient mechanisms for data collection and transmission are of paramount importance. A popular technique for achieving energy efficiency is Compressive Sensing (CS). CS's major advantage is that it can sample and compress information in a single step. At the same time, CS offers lightweight data encryption during compression. In this work, we propose an adaptive CS scheme where a central SO with enhanced capabilities utilizes a learning phase, associating different sparsity levels with the reconstruction error. Based on this scheme, this SO provides feedback to the rest of the SOs to adapt their CS parameters in order to serve applications with a diverge range of quality-of-service requirements.

Infrastructures enabling experimental assessment of Internet of Things (IoT) solutions are scarce. Moreover, such infrastructures are typically bound to a specific application domain, thus, not facilitating the testing of solutions with a horizontal approach. This paper presents a platform that supports Experimentation as s Service (EaaS) over a federation of IoT testbeds. This platform brings two major advances. First, it leverages semantic web technologies to enable interoperability so that testbed agnostic access to the underlying facilities is allowed. Second, a set of tools ease both the experimentation workflow and the federation of other IoT deployments, independently of their domain of interest. Apart from the platform specification, this paper presents how this design has been actually instantiated into a cloud-based EaaS platform that has been used for supporting a wide variety of novel experiments targeting different research and innovation challenges. In this respect, this paper summarizes some of the experiences from these experiments and the key performance metrics that this instance of the platform has exhibited during the experimentation.

One of the main features of future networks will be heterogeneity in the wireless access environment in which a mobile terminal will be able to connect to multiple radio networks simultaneously. Network selection and efficient load balancing among different types of networks will be required to achieve high-speed connectivity with seamless mobility. Game theory is a mathematical tool developed to understand competitive situations in which rational decision makers interact to achieve their objectives. Game theory techniques can be easily adapted for use in radio resource management mechanisms in a heterogeneous environment. To this end, we propose here a theoretical scheme for fusing network selection mechanisms and game theory, so that access admission control can be modeled efficiently as a non-cooperative game. In this sense, networks play against each other so as to maximize their payoff and admission control policy ensures maximum QoS for all service requests.