We address the problem of characterising the aggregate flexibility in populations of electric vehicles (EVs) with uncertain charging requirements. Building on previous results that provide exact characterisations of the aggregate flexibility in populations with known charging requirements, in this paper we extend the aggregation methods so that charging requirements are assumed to be uncertain, but sampled from a known distribution. In doing so we, construct distributionally robust aggregate flexibility sets, sets of aggregate charging profiles, over which we can provide probabilistic guarantees, that sampled populations will be able to track. By leveraging certain measure concentration results that establish powerful finite sample guarantees, we are able to give tight bounds on these robust flexibility sets, even in low sample regimes that are well suited for aggregating small populations of EVs. We detail explicit methods of calculating these sets and provide numerical results that validate the theory developed here.

Mixture of Experts (MoE) is an advanced machine learning paradigm that enhances model scalability, efficiency, and specialization by dynamically selecting subsets of experts for each input. This survey provides a comprehensive overview of MoE, covering its fundamental principles, advantages, challenges, recent advancements, and future research directions. MoE architectures have been widely adopted in large-scale artificial intelligence systems, particularly in natural language processing and computer vision, due to their ability to optimize computational resources while maintaining high model capacity. Key challenges such as training instability, expert imbalance, computational overhead, and interpretability are discussed alongside potential mitigation strategies. Recent innovations, including hierarchical MoE, improved gating mechanisms, adaptive expert pruning, and MoE integration into multimodal learning, have significantly enhanced the framework's effectiveness. The survey also explores future research directions, focusing on improving training stability, fairness, continual learning, and resource efficiency. As MoE continues to evolve, it remains a promising approach for developing highly efficient and scalable AI systems.

Coastal regions are extremely vulnerable to storms and tropical cyclones, which have caused significant economic costs and numerous fatalities. This underscores the urgent need for action to protect the sustainability and resilience of coastal communities. In this paper, we present an AI-driven geospatial analysis pipeline for automating coastal disaster assessment by detecting building damage from satellite imagery. Firstly, we propose an effective and scalable pipeline to train an artificial intelligence (AI) model for damaged building detection using a limited dataset. Specifically, we pretrain our AI model with Microsoft's Building Footprint dataset to adapt to Puerto Rico's landscape, then fine-tune it using a mix of manually and self-annotated data. Upon training, we used our AI model to generate geospatial heatmaps of damaged building counts and damage ratio, which are useful to assess the storm damage and coastal vulnerability. Our approach enhances disaster response efficiency and supports better planning and resilience strategies. Additionally, it can significantly reduce labour costs associated with manual ground surveys for insurance claims and post-disaster socioeconomic impact evaluation. Our solution is one of the Top 10 Global Semi-Finalists in 2024 EY Open Science Data Challenge.

Decode-and-forward and amplify-and-forward relay attacks are a powerful tool for defeating challenge-response authentication protocols. Current solutions for detecting these relay attacks utilize round-trip time distance-bounding. Unfortunately, secure implementations of distance-bounding require dedicated ultra-wideband hardware that only provide low data rates and operate at relatively short distances. In this paper we propose two novel symmetric-key challenge-response authentication protocols that can detect either a decode-andforward relay attack or prevent a decode-and-forward and detect an amplify-and-forward relay attack. Both protocols utilize the channel state information in a far-field communication system to perform the detection. The first protocol utilizes the correlation of the adjusted channel frequency response to detect decode-and-forward relay attacks. The second protocol prevents decode-and-forward relay attacks through the use of randomized pilots, and detects amplify-and-forward relay attacks by classifying the distribution of the channel frequency response that is caused by multiple relays. The protocols utilize orthogonal frequency division multiplexing to estimate the channel frequency response between the legitimate communicating parties to identify if a relay attack is occurring in a physical-layer challenge-response authentication protocol. The proposed protocols can be leveraged on many existing hardware platforms and can simultaneously support high data rates. To evaluate the performance of the protocol, MATLAB simulations are done to gather Monte-Carlo results on performance criterion.

Data science is revolutionizing academia and industry, and there is a strong demand for a workforce fluent in data science. The availability of such courses has increased substantially in recent years. However, there are few rigorous curricula built on mathematical logic as the foundation of data science. The LogicDS Project aims to engage high school students from rural communities in an online data science course that integrates mathematics, statistics, and programming concepts into a single unified framework based on logic and reasoning. We developed a one-week data science course consisting of six lessons (a sampling of the full course offering) and recruited 110 participants. We collected pre- and post-intervention data and students' LMS activity log data to analyze their engagement. Results indicate that our Logic-Based framework for data science education effectively engages students from a variety of backgrounds; it was shown that they perceived the course as valuable for learning data science skills/concepts. Notably, our focus on the entropy analysis of student activity logs correlated to our other mixed methods analyses. This study provided insights for engaging K-12 students learning data science.

The controlled waterway in the upper reaches of the Yangtze River has become a bottleneck for shipping due to its curved, narrow and turbulent characteristics. The vessels passing through it must obey the signal revealed by the Intelligent Vessel Traffic Signaling System (IVTSS) to pass in one direction. The accuracy of signals is directly related to traffic safety and efficiency. However, the unreliability of vessel sensing sensors in these areas and the latency of transmission and computation of large amounts of sensing data may negatively impact IVTSS. Hence, more information from the physical world is needed to ensure the stable operation of the IVTSS, and we proposed an edge computing-centric sensing and execution system based on IoT architecture to enhance the reliability of IVTSS. We conducted experiments using plug-and-play methods, reducing command and recording error rates by 89.47% and 86.27%, respectively, achieving the goal of real-time perception control.

This paper presents a feature selection technique based on ℓ1 regularization to select the most relevant weights of artificial neural networks (ANNs) for digital predistortion (DPD) linearization of wideband radio-frequency (RF) power amplifiers (PAs). The proposed pruning method is applied to the first hidden layer of a feed-forward real-valued time-delay neural network, commonly used for DPD purposes. In addition, this paper presents the ANN-based DPD implementation using a graphic processing unit (GPU) with compute unified device architecture (CUDA) units. Thanks to the proposed pruning strategy, it is possible to reduce the ANN complexity significantly, thereby achieving a higher data throughput with the GPUbased implementation. The trade-off among RF performance metrics, number of model parameters and throughput of the GPU implementation is evaluated considering the linearization of a high-efficiency pseudo-Doherty load modulated balanced amplifier (LMBA). The linearized PA operating at 2 GHz RF frequency delivers a mean output power of 40 dBm with around 50% power efficiency when being excited with 5G new radio signals with up to 200 MHz bandwidth and 8 dB peak-to-average power ratio (PAPR). The real-time GPU implementation of the ANN-based DPD can meet the linearity specifications with a throughput circa 1 GSa/s.

This study presents a computational simulation to explore the distinctive patterns of neural connectivity in Autism Spectrum Disorder (ASD), focusing on the concepts of spatial and topological proximities. ASD is characterized by unique cognitive and behavioral profiles, often attributed to atypical neural development and connectivity. Our simulation employs a two-pronged approach: first, by modeling spatial proximity through the generation of clustered 3D node positions, we mimic the observed over-connectivity in localized brain regions in ASD. Second, we construct a topological network to represent the complex interplay of under-connectivity and over-connectivity in functional brain networks, a hallmark of ASD's neural landscape. The spatial model demonstrates pronounced clustering, reflecting the structural brain differences and atypical growth patterns reported in neuroimaging studies of individuals with ASD. In contrast, the topological model reveals a network with a mix of sparse and dense connections, simulating the diverse and often contradictory findings of functional connectivity in ASD, such as reduced long-range connections and increased local connections. Our simulation provides a visual and conceptual framework for understanding the altered neural connectivity in ASD. It underscores the importance of considering both physical (spatial) and functional (topological) aspects of brain connectivity to grasp the full extent of neurodevelopmental deviations in ASD. This approach not only aids in elucidating the neural underpinnings of ASD's core symptoms but also offers a foundation for developing targeted interventions. Future research, integrating more complex and individual-specific data, could further refine this model, enhancing our understanding of ASD's neural dynamics and contributing to personalized therapeutic strategies.

We study burst erasure correcting streaming codes for three-node relay networks, where there is a source-relay link and a relay-destination link. These codes guarantee that all message packets are recovered within a delay of τ time slots, given that a single burst erasure of length at most b packets occurs in both links. Leveraging previously known techniques in the streaming code literature, we first provide a simple upper bound on the rate of burst erasure correcting streaming codes for three-node relay networks. Our main result is a coding scheme that achieves rates arbitrarily close to the rate upper bound, as message size increases.

The intersection of robotics and artificial intelligence led to a profound paradigm shift in Robot Learning. Robots have the capacity to replicate human actions and also dynamically adapt, innovate, and excel across a spectrum of tasks. However, the heterogeneity in the deployment of robot platforms and software frameworks poses considerable challenges in terms of systematic testing and comparative analyses. Additionally, the data scarcity of especially force controlled robot manipulation is still restraining the development of advanced foundation models. A reference platform with default software stack can help to increase comparability, reducing development time and collect a large amount of tactile robot manipulation data. To address on this problem, we developed a Parallel and Distributed Robot AI (PD.RAI) framework, comprising a scalable ensemble of Robot Learning Units (RLUs), a global database, and the Robot Cluster Intelligence (RoCI). Each RLU is endowed with robot arms, cameras, and local computational units to autonomously engage in planning, control, and local machine learning of tactile manipulation skills. The RoCI system oversees the learning process and schedules the RLUs tasks. To show the functionality of the system, two black-box optimization algorithms are compared within the robot skill learning domain. An experiment with 24 different optimization tasks is conducted in parallel. The algorithms are incorporated into the same existing default modules acting as a reference environment. This allows for a realistic comparison without sacrificing diversity of possible configurations and testing environments.

Raven's Progressive Matrices (RPMs) have been widely used for measuring abstract reasoning and intelligence in humans. However for artificial learning systems, abstract reasoning remains a challenging problem. This paper investigates the potential of biologically inspired spiking neural networks in addressing this challenge. We focus on unsupervised learning since in most domains where human-competitive conceptual learning is important, unlabelled data is plentiful and the autonomous discovery and understanding of underlying patterns without explicit guidance is paramount. Our experiments show that the overall accuracy of our unsupervised networks significantly outperform some supervised methods. Our results also demonstrate that, unlike their non-spiking counterparts, spiking neural networks are able to extract and encode relational features without any explicit instruction, do not rely on labelled training data, and are able to simulate the human ability of navigating and deciphering complex data landscapes. Our results exhibit state-of-the-art performance for unsupervised learning on the RAVEN (67.0%) and I-RAVEN (40.3%) datasets, indicating that that spiking neural networks are well suited to unsupervised conceptual learning.

Telecom networks are becoming increasingly complex, with diversified deployment scenarios, multi-standards, and multi-vendor support. The intricate nature of the telecom network ecosystem presents challenges to effectively manage, operate, and optimize networks. To address these hurdles, Artificial Intelligence (AI) has been widely adopted to solve different tasks in telecom networks. However, these conventional AI models are often designed for specific tasks, rely on extensive and costlyto-collect labeled data that require specialized telecom expertise for development and maintenance. The AI models usually fail to generalize and support diverse deployment scenarios and applications. In contrast, Foundation Models (FMs) show effective generalization capabilities in various domains in language, vision, and decision-making tasks. FMs can be trained on multiple data modalities generated from the telecom ecosystem and leverage specialized domain knowledge. Moreover, FMs can be fine-tuned to solve numerous specialized tasks with minimal task-specific labeled data and, in some instances, are able to leverage context to solve previously unseen problems. At the dawn of 6G, this paper investigates the potential opportunities of using FMs to shape the future of telecom technologies and standards. In particular, the paper outlines a conceptual process for developing Telecom FMs (TFMs) and discusses emerging opportunities for orchestrating specialized TFMs for network configuration, operation, and maintenance. Finally, the paper discusses the limitations and challenges of developing and deploying TFMs.

A document by Luis Hernandez Corporales . Click on the document to view its contents.

Successive Interference Cancellation (SIC) is a key element in a Non-Orthogonal Multiple-Access (NOMA) receiving terminal and must be properly tuned in order to achieve the optimal performance. To this purpose, receivers have to be informed about which signal components have to be decoded before decoding their own. In this framework, most literature refers to a proposition reported in [1, Summary 6.1] for broadcast Gaussian channels implying that the optimal decoding order consists of decoding the weaker user signals first and then their own. In this work, the blurred contours of the proposition are brought into focus by considering the widespread Proportional Fairness (PF) criterion to harmonize the user rates. Theorem 1 is the main contribution assessing the optimal SIC order of a NOMA channel under a PF-optimal criterion. It is observed in the paper that the optimum SIC order may change with different utility functions which are concave and monotonically increasing with the users' individual SINRs. Numerical results are also presented to confirm the validity in a fading environment.

In massive multiple-input multiple-output (MIMO) systems, cellular users are affected by inter-cell coherent interference, especially pilot contamination. To solve this problem and improve spectral efficiency (SE) of each cell, we propose a reconfigurable intelligent surface (RIS)-assisted cellular massive MIMO system. Based on the uplink pilot information and minimum mean square error (MMSE) method, base stations (BSs) can estimate all the uplink channels for local decoding. Then, we give an uplink signal-to-noise-plus-interference ratio (SINR) expression and an each user's SE expression based on maximum ratio combining (MRC) processing method. The effects of pilot contamination, beamforming gain, interference, and noise on the system can be seen from the expressions. Power control and RIS phase shift optimization methods are also proposed to reduce pilot contamination. The power control method is formulated to maximize each user's rate. The phase shift optimization is also formulated to maximize the target area's average power. Numerical results show that power control method and phase shift matrix optimization can improve the system performance.

A document by Florian Reißner . Click on the document to view its contents.

Soccer is one of the most popular sports worldwide, with millions of fans and players. Teams often invest significant resources into improving their performance and achieving success in their respective leagues. One key aspect of a team’s performance is their playing style, which reflects their strengths and weaknesses on the field. Understanding how playing styles vary across different countries and leagues can provide valuable insights for coaches, managers, and analysts. Understanding how playing styles vary across different countries and leagues can provide valuable insights for coaches, managers, and analysts. In this project, we aim to analyze the playing styles of professional soccer teams from six different countries: Italy, Germany, Spain, England, France, and the United States. By using K-means clustering to group teams, we can understand how their playing styles differ from each other. We will use real datasets from the previous year and compare our results with past research papers to provide insights into the similarities and differences in playing styles across different countries and leagues. Our findings will inform coaching strategies and improve our understanding of the game.

Meeting the targets of Sustainable Development Goal (SDG) 7, which focuses on ensuring access to affordable, reliable, sustainable, and modern energy for all, poses significant challenges. Overcoming these hurdles requires innovative solutions that can bridge the gap between current capabilities and future needs. Swarm electrification emerges as a promising concept that could accelerate progress towards achieving SDG 7 goals by leveraging the collective power of decentralized energy resources. This review delves into the concept of swarm electrification, placing it within the context of the prevailing trends in the power system sector: decentralization, decarbonization, and digitalization. It examines the role of digital technologies in enhancing swarm electrification and categorizes application areas according to the phases of swarm electrification. Particular attention is given to the technologies underpinning Deep Digitalization, such as distributed ledger technology, notably blockchain, and artificial intelligence, with a focus on machine learning. These technologies play pivotal roles in advancing swarm electrification. The review demonstrates how deep digitalization can facilitate the improvement of swarm electrification and ultimately support the integration of bottom-up initiatives with top-down grid expansion efforts over time.