Overfitting is a well-known issue in machine learning that occurs when a model struggles to generalize its predictions to new, unseen data beyond the scope of its training set. Traditional techniques to mitigate overfitting include early stopping, data augmentation, and regularization. In this work, we demonstrate that the degree of overfitting of a trained model is correlated with the ability to generate counterfactual examples. The higher the overfitting, the easier it will be to find a valid counterfactual example for a randomly chosen input data point. Therefore, we introduce CF-Reg, a novel regularization term in the training loss that controls overfitting by ensuring enough margin between each instance and its corresponding counterfactual. Experiments conducted across multiple datasets and models show that our counterfactual regularizer generally outperforms existing regularization techniques.

Explainable Artificial Intelligence (XAI) has emerged as a critical area of research to unravel the opaque inner logic of (deep) machine learning models. Among the various XAI techniques proposed in the literature, counterfactual explanations stand out as one of the most promising approaches. However, these “what-if” explanations are frequently complex and technical, making them difficult for non-experts to understand and, more broadly, challenging for humans to interpret. To bridge this gap, in this work, we exploit the power of open-source Large Language Models to generate natural language explanations when prompted with valid counterfactual instances produced by state-of-the-art explainers for graph-based models. Experiments across several graph datasets and counterfactual explainers show that our approach effectively produces accurate natural language representations of counterfactual instances, as demonstrated by key performance metrics

During advanced surveillance missions, Unmanned Aerial Vehicles (UAVs) usually require the execution of edge-assisted computer vision (CV) tasks. In multi-hop UAV networks, the successful transmission of these tasks to the edge is severely challenged due to severe bandwidth constraints, and the possible node failures. To address these critical challenges, we propose a novel A 2 - UAV framework that optimizes the number of correctly executed tasks at the edge. In stark contrast with existing art, we take an application-aware approach and formulate a novel Application-Aware Task Planning Problem ( A 2 - TPP ) to optimize routing, data pre-processing and target assignment for each UAV. Our formulation explicitly takes into account (i) the relationship between CV task accuracy and image compression for the classes of interest based on the available dataset, (ii) the target positions , (iii) the current energy/position of the UAVs, and (iv) the possible node failures. We demonstrate A 2 - TPP is NP-Hard and propose a polynomial-time algorithm to solve it efficiently. We extensively evaluate A 2 - UAV through simulation and real-world experiments using a testbed composed by four DJI Mavic Air 2 UAVs. Results on image classification show that A 2 - UAV attains on average around 38% more accomplished tasks w.r.t. the state of the art, with a 400% improvement in tasks-intensive scenarios. Moreover, we show that our framework is able to reconfigure the network in case of nodes failure.

To perform advanced surveillance, Unmanned Aerial Vehicles (UAVs) require the execution of edge-assisted computer vision (CV) tasks. In multi-hop UAV networks, the successful transmission of these tasks to the edge is severely challenged due to severe bandwidth constraints. For this reason, we propose a novel A2-UAV framework to optimize the number of correctly executed tasks at the edge. In stark contrast with existing art, we take an application-aware approach and formulate a novel Application-Aware Task Planning Problem (A2-TPP) that takes into account (i) the relationship between deep neural network (DNN) accuracy and image compression for the classes of interest based on the available dataset, (ii) the target positions, (iii) the current energy/position of the UAVs to optimize routing, data pre-processing and target assignment for each UAV. We demonstrate A2-TPP is NP-Hard and propose a polynomial-time algorithm to solve it efficiently. We extensively evaluate A2-UAV through real-world experiments with a testbed composed by four DJI Mavic Air 2 UAVs. We consider state-of-the-art image classification tasks with four different DNN models (i.e., DenseNet, ResNet152, ResNet50 and MobileNet-V2) and object detection tasks using YoloV4 trained on the ImageNet dataset. Results show that A2-UAV attains on average around 38% more accomplished tasks than the state of the art, with 400% more accomplished tasks when the number of targets increase significantly. To allow full reproducibility, we pledge to share datasets and code with the research community.

Swarms of Unmanned Aerial Vehicles (UAVs) are a key technology to support communication in many harsh environments where fixed infrastructures (e.g., 5G) are disrupted or not available. However, the fast mobility and highly dynamic network topology pose unique challenges and require the development of novel multi-hop routing protocols. Previous work in this direction extends geographical protocols or adapts approaches designed for Mobile Ad-hoc NETworks (MANETs), rarely taking full advantage of UAV capabilities. In this paper, we introduce a novel data offloading approach, namely Stop & Offload, that exploits the device controllable mobility to facilitate network routing. The swarm of UAVs performs data offloading synchronously and recurrently. At fixed intervals of time, the swarm interrupts the sensing mission (Stop) and moves, as little as possible, to build a connected formation to the base station and offload the data (Offload). We provide both centralized solutions — assuming a long-range control channel — and a distributed solution — working in the absence of a control channel. By means of extensive simulations we show that our proposals outperform state-of-the-art solutions, decreasing the time taken to build a connected formation of about 45% and increasing the time spent on sensing of 10%. Additionally, we compared our protocol with various routing strategies and observe remarkable improvements, including a 50% reduction in average packet delay.