Advertising click-through rate (CTR) prediction is a fundamental task in recommender systems, aimed at estimating the likelihood of users interacting with advertisements based on their historical behavior. This prediction process has evolved through two main stages: from traditional shallow interaction models to more advanced deep learning approaches. Shallow models typically operate at the level of individual features, failing to fully leverage the rich, multilevel information available across different feature sets, leading to less accurate predictions. In contrast, deep learning models exhibit superior feature representation and learning capabilities, enabling a more realistic simulation of user interactions and improving the accuracy of CTR prediction. This paper provides a comprehensive overview of CTR prediction algorithms in the context of recommender systems. The algorithms are categorized into two groups: shallow interactive models and deep learning-based prediction models, including deep neural networks, convolutional neural networks, recurrent neural networks, and graph neural networks. Additionally, this paper also discusses the advantages and disadvantages of the aforementioned algorithms, as well as the benchmark datasets and model evaluation methods used for CTR prediction. Finally, it identifies potential future research directions in this rapidly advancing field.

This paper introduces a distributed online learning coverage control algorithm based on sparse Gaussian process regression for addressing the problem of multi-robot area coverage and source localization in unknown environments. Considering the limitations of traditional Gaussian process regression in handling large datasets, this study employs multiple robots to explore the task area to gather environmental information and approximate the posterior distribution of the model using variational free energy methods, which serves as the input for the centroid Voronoi tessellation algorithm. Additionally, taking into consideration the localization errors, and the impact of obstacles, buffer factors and centroid Voronoi tessellation algorithms with separating hyperplanes are introduced for dynamic robot task area planning, ultimately achieving autonomous online decision-making and optimal coverage. Simulation results demonstrate that the proposed algorithm ensures the safety of multi-robot formations, exhibits higher iteration speed, and improves source localization accuracy, highlighting the effectiveness of model enhancements.

DLV2 is an AI tool for knowledge representation and reasoning that supports answer set programming (ASP) – a logic-based declarative formalism, successfully used in both academic and industrial applications. Given a logic program modeling a computational problem, an execution of DLV2 produces the so-called answer sets that correspond one-to-one to the solutions to the problem at hand. The computational process of DLV2 relies on the typical ground & solve approach, where the grounding step transforms the input program into a new, equivalent ground program, and the subsequent solving step applies propositional algorithms to search for the answer sets. Recently, emerging applications in contexts such as stream reasoning and event processing created a demand for multi-shot reasoning: here, the system is expected to be reactive while repeatedly executed over rapidly changing data. In this work, we present a new incremental reasoner obtained from the evolution of DLV2 toward iterated reasoning. Rather than restarting the computation from scratch, the system remains alive across repeated shots, and it incrementally handles the internal grounding process. At each shot, the system reuses previous computations for building and maintaining a large, more general ground program, from which a smaller yet equivalent portion is determined and used for computing answer sets. Notably, the incremental process is performed in a completely transparent fashion for the user. We describe the system, its usage, its applicability, and performance in some practically relevant domains.

Ideological and relational polarization are two increasingly salient political divisions in Western societies. We integrate the study of these phenomena by describing society as a multilevel network of social ties between people and attitudinal ties between people and political topics. We then define and propose a set of metrics to measure ‘network polarization’: the extent to which a community is ideologically and socially divided. Using longitudinal network modelling, we examine whether observed levels of network polarization can be explained by three processes: social selection, social influence, and latent-cause reinforcement. Applied to new longitudinal friendship and political attitude network data from two Swiss university cohorts, our metrics show mild polarization. The models explain this outcome and suggest that friendships and political attitudes are reciprocally formed and sustained. We find robust evidence for friend selection based on attitude similarity and weaker evidence for social influence. The results further point to latent-cause reinforcement processes: (dis)similar attitudes are more likely to be formed or maintained between individuals whose attitudes are already (dis)similar on a range of political issues. Applied across different cultural and political contexts, our approach may help to understand the degree and mechanisms of divisions in society.

How areas of land are allocated for different uses, such as forests, urban areas, and agriculture, has a large effect on the terrestrial carbon balance and, therefore, climate change. Based on available historical data on land-use changes and a simulation of the associated carbon emissions and removals, a surrogate model can be learned that makes it possible to evaluate the different options available to decision-makers efficiently. An evolutionary search process can then be used to discover effective land-use policies for specific locations. Such a system was built on the Project Resilience platform and evaluated with the Land-Use Harmonization dataset LUH2 and the bookkeeping model BLUE. It generates Pareto fronts that trade off carbon impact and amount of land-use change customized to different locations, thus providing a proof-of-concept tool that is potentially useful for land-use planning.

The autonomous safe flight of fixed-wing unmanned aerial vehicles (UAVs in complex low-altitude environments presents significant challenges and holds practical application value. This paper proposes a motion planning method for agile fixed-wing UAVs to address safety issues in navigating narrow corridors within such environments. In the path planning phase, we introduce the Improved Batch Informed Trees (IBIT*) to enhance both the solving speed and quality of BIT*. The IBIT* incorporates strategies such as using Rapidly Exploring Random Tree (RRT)-Connect for initial pathfinding, informed sparse sampling, and re-selecting parent nodes. During the trajectory planning phase, we first decouple the roll angle of the UAV from its three-dimensional position based on the agility of fixed-wing UAVs; subsequently, we address constraints related to smoothness and mission time by leveraging the characteristics of the Minimum Control Effort; finally, we design a differentiable penalty function to satisfy the dynamic performance constraints of the UAV. The effectiveness and superiority of the proposed motion planning method are demonstrated through numerical simulations and physical flight experiments.

This article explores how online language learners encounter foreign language speaking anxiety (FLSA), what mitigating strategies they apply to manage synchronous online tutorials, and what their asynchronous speaking practices are. In a large-scale mixed methods study, we gathered survey data from 307 language learners at a UK online and distance learning university and conducted in-depth group interviews with 10 students focusing on their FLSA experience and perceptions regarding synchronous and asynchronous speaking activities. The results reveal that the triggers of FLSA and the mitigating strategies learners apply partly overlap with those in the face-to-face context but are partly specific to the online environment (e.g. breakout rooms, vicarious learning). The use of technology can be anxiety-inducing (e.g. cameras) as well as supportive (e.g. online translation tools and dictionaries). Novel findings of the study are that avoidance strategies are more nuanced in this context, ranging from complete avoidance of tutorials to full engagement via the chat, and that the use of breakout rooms magnifies learners’ emotions and is one of the main triggers of FLSA. This might be helpful for practitioners – also beyond language courses – in scaffolding and optimising their small group activities online.

Outdoor mobile robots must navigate uneven terrains with obstacles that sometimes cannot be avoided; therefore, strategies have been developed for robots to overcome them. In most cases, these strategies have been modeled considering movement over horizontal surfaces and with the robot positioned directly in front of the obstacle, but this idealization does not occur in most real cases. Therefore, this article describes a strategy for obstacle overcoming, useful when the robot faces obstacles on inclined terrains and at an oblique angle relative to the robot’s trajectory, considering rollover stability during the process, based on the reaction force criterion. This strategy can be used by mobile robots with wheels and an articulated arm whose end effector can contact the ground, and it consists of a sequence of standard movements that include the use of the arm, whose variable location was defined through a system developed using fuzzy logic. The designed strategy was validated through simulations and then implemented on the Lázaro robot, verifying its effectiveness through experimental tests. With it, the robot can overcome obstacles such as steps, ramps, and ditches from any position; additionally, it increased the ability to overcome obstacles with a height close to twice the radius of the robot’s wheels.

Parallel manipulators (PMs) are adopted in different fields due to their superior characteristics compared to serial manipulators. PMs with flexible links are likely more energy efficient and have high dynamic performance since they are lighter than those with rigid links. On the other hand, due to their lightweight design, the flexibility can lead to undesired deformation and vibration, decreasing the tracking trajectory and transient errors. This work proposes a two-loop active vibration control strategy, using strain gauges and piezoelectric lead zirconate (PZT) actuators, to compensate for the undesired effect of the flexibility. A pose control loop exploits the sliding mode control using data collected from images acquired by an oCam-5CRO-U camera, while the active vibration control loop uses strain gauge sensors and PZT actuators. Strain gauges are responsible for measuring the deformation of each link, and after being treated by digital filters, these signals are applied to the PZT actuator. Combining both loops allows the manipulator to be guided over the desired trajectory with positive vibration attenuation. The results reveal that the presence of the PZT on both sides of the flexible links increases the links’ rigidity, yielding overshoot and vibration reduction during the manipulator’s motion. In addition, the maximum peak is significantly attenuated, and the overall oscillations are also positively reduced when using the two-loop active control strategy. The root mean square error quantifies this attenuation, showing an average reduction of 30% in the corresponding step input directions. Therefore, the proposal improves the system performance by enhancing the tracking trajectory with lower vibrations.

Locomotion control of inchworm robots presents significant challenges due to their highly nonlinear dynamics and complex interactions with the environment. Traditional control methods often struggle with achieving precise tracking and adaptive performance in dynamic conditions. To address these limitations, this article proposes a novel data-driven compound control system that integrates fractional proportional-integral derivative (FPID) control with Koopman operator theory. Unlike conventional approaches, which rely on direct nonlinear control or simplified linear approximations, our method leverages data-driven modeling to transform the nonlinear dynamics into a linear representation, making control design more systematic and scalable. A deep neural network is trained to identify the Koopman operator, enabling an FPID controller to operate within this transformed space for improved tracking accuracy and robustness. The proposed framework is validated using NVIDIA Isaac SIM simulation software, demonstrating superior locomotion efficiency and tracking performance compared to existing control strategies. This study advances the control of bio-inspired robots by bridging fractional-order control with data-driven Koopman-based modeling, addressing the fundamental challenge of achieving high-precision locomotion in complex environments.

Connecting individual robots to form an inter-reconfigurable system with a flexible base size enhances the ability to access and cover areas for cleaning and maintenance tasks. Given that increased configuration complexity expands the search space dimension, an optimal routing solution ensuring efficiency is essential. In this paper, we present an inter-reconfigurable multi-robot system capable of adjusting the bases of its two units, along with an optimal path planning approach for confined spaces based on a modified informed rapidly-exploring random tree algorithm by a greedy set (RIRRT*). We validate the navigation of the proposed inter-reconfigurable platform using RIRRT* for four informed dimensional search spaces as a case study in both simulated and real-world environments. The proposed path planning method for the inter-reconfigurable system outperformed conventional strategies, achieving significant reduction in both execution time and energy utilization.