Recommendation systems chiefly rely on Collaborative Filtering techniques to depict user inclinations by analyzing historical interactions. This traditional approach has seen significant improvements with recent advan...
ISBN:
(纸本)9789819601158;9789819601165
Recommendation systems chiefly rely on Collaborative Filtering techniques to depict user inclinations by analyzing historical interactions. This traditional approach has seen significant improvements with recent advancements that leverage Graph Neural Networks (GNNs). GNNs enhance CF by exploiting high-order interactions within user-item bipartite graphs [1]. However, existing GNN-based CF models face a critical limitation: the entanglement of depth and range, where increasing GNN layers simultaneously enlarges the receptive field, complicating the model and increasing computational demands. To address this, we propose a framework that decouples range from depth, using Personalized PageRank (PPR) to extract localized subgraphs with bounded ranges, allowing for deeper GNN layers without expanding the receptive field. Additionally, we introduce a learnable aggregation function that optimally integrates sum, max, and mean operations for improved message aggregation. Experiments on multiple datasets demonstrate our approach's superior performance. The results indicate that our framework not only addresses the depth and range entanglement issue but also significantly boosts the effectiveness of GNN-based CF models.
The ever-expanding volume of data presents considerable challenges in storing and processing semi-supervised models, hindering their practical implementation. Researchers have explored reducing network versions as a p...
ISBN:
(纸本)9783031790287;9783031790294
The ever-expanding volume of data presents considerable challenges in storing and processing semi-supervised models, hindering their practical implementation. Researchers have explored reducing network versions as a potential solution. Real-world networks often comprise diverse vertex and edge types, leading to the adoption of k-partite network representation. However, existing methods have mainly focused on reducing uni-partite networks with a single vertex type and edges. This study introduces a novel coarsening method designed explicitly for k-partite networks, aiming to preserve classification performance while addressing storage and processing issues. We conducted empirical analyses on synthetically generated networks to evaluate their effectiveness. The results demonstrate the potential of coarsening techniques in overcoming storage and processing challenges posed by large networks. The proposed coarsening algorithm significantly improved storage efficiency and classification runtime, even with moderate reductions in the number of vertices. This led to over one-third savings in storage space and a twofold increase in classification speed. Moreover, the classification performance metrics exhibited low variation on average, indicating the algorithm's robustness and reliability in various scenarios.
Opinion polarization and political segregation are key societal concerns, especially on social media. Although these phenomena have been traditionally attributed to homophily-preference for like-minded individuals-rec...
ISBN:
(纸本)9783031789793;9783031789809
Opinion polarization and political segregation are key societal concerns, especially on social media. Although these phenomena have been traditionally attributed to homophily-preference for like-minded individuals-recent work in social psychology suggests that acrophily-preference for extreme rather than moderate opinions-might play a role as well. In this work, we introduce a methodology to estimate the degree of preference for connecting with users who hold strong opinions on social media. Our framework is composed of four phases: (i) opinion estimation, (ii) opinion thresholding, (iii) network construction, and (iv) acrophily estimation. We apply it to study the climate change debate on Reddit and find that users show higher-than-expected acrophilic patterns, especially if they are climate skeptics or have extreme opinions. Acrophilic patterns are stable over time, while polarization gradually leaves space for pluralism.
Challenging weather conditions such as fog pose a big problem in vehicle detection because it hinders easy recognition and affects the quality of the image even more. In this work, vehicle detection is investigated wi...
ISBN:
(纸本)9789819607884;9789819607891
Challenging weather conditions such as fog pose a big problem in vehicle detection because it hinders easy recognition and affects the quality of the image even more. In this work, vehicle detection is investigated with an enhancement of accuracy for poor lighting conditions through the employment of YOLO deep learning algorithm coupled with Retinex image enhancement technique. YOLO is an advanced object detection algorithm that is fast and very precise but is not sensitive to low light areas. Likely to improve the image visibility and detection reliability by mimicking mechanisms of a human eye, the Retinex algorithm is capable to mitigate the negative impact of fog. The overall idea presented in this paper is based on the pipelined approach, where Retinex algorithm will be employed initially to improve the input images to be fed to the YOLO vehicle detection system. The effectiveness of this approach is tested using a set of foggy weather images revealing much higher level of detection performance and precision as compared to YOLO alone.
ErgoAI (abbr. Ergo) is a high level, multi-paradigm logic programming language and system developed by Coherent Knowledge Systems as an enhancement of and a successor to the popular Flora-2 system. Ergo is oriented to...
ISBN:
(纸本)9783031742088;9783031742095
ErgoAI (abbr. Ergo) is a high level, multi-paradigm logic programming language and system developed by Coherent Knowledge Systems as an enhancement of and a successor to the popular Flora-2 system. Ergo is oriented towards scalable knowledge representation and reasoning, and can exploit both structured knowledge as well as knowledge derived from external sources such as vector embeddings. From the start, Ergo (and Flora-2 before it) were designed to exploit the well-founded semantics for reasoning in a multi-paradigm environment, including object-based logic (F-logic) with non-monotonic inheritance;higher order syntax in the style of HiLog;defeasibility of rules;semantically clean transactional updates;and extensive use of subgoal delay for better logical behavior and performance. Although Ergo programs are compiled into XSB and adopt many Prolog features, Ergo is altogether a different language with functionality equivalent to major Prologs.
This paper introduced a three-degree-of-freedom (3-DOF) photoelectric stabilized platform (PSP), and the Finite-Step-Integration (FSI) method was applied to analyze the kinematics characteristics. The motion process o...
ISBN:
(纸本)9789819607709;9789819607716
This paper introduced a three-degree-of-freedom (3-DOF) photoelectric stabilized platform (PSP), and the Finite-Step-Integration (FSI) method was applied to analyze the kinematics characteristics. The motion process of the mechanism is segmented linearly, and the mathematical model of forward kinematics (FK) with iterative parameters is created and solved. The kinematic simulation is verified in ADAMS software, and the calculated results are compared with the simulation. The workspace of the PSP is analyzed to describe the working dimensionality characteristics. The results show that: the FK algorithm based on the FSI method has enough computational precision and can be efficient for the workspace analysis. This paper's research provides a good reference for the design and kinematics characteristic analysis of PSP.
This article proposes a novel two-layer mission planning structure for multi-robot collaborative area coverage mission planning: (1) First, an improved co-evolving particle swarm optimization algorithm is utilized to ...
ISBN:
(纸本)9789819607884;9789819607891
This article proposes a novel two-layer mission planning structure for multi-robot collaborative area coverage mission planning: (1) First, an improved co-evolving particle swarm optimization algorithm is utilized to resolve the virtual sensor configuration problem, obtaining a sensor configuration scheme that can fully cover the searchable mission area;(2) Using an improved K-means algorithm to cluster and partition the sensor configuration points, obtaining each sub-area and the search path within each sub-area. The simulation results confirm the effectiveness of the method proposed in this article.
Social interactions are shaped by homophily, the tendency for individuals to connect with others who share similar attributes. Exploring this phenomenon is crucial for understanding a wide spectrum of social behaviors...
ISBN:
(纸本)9783031789793;9783031789809
Social interactions are shaped by homophily, the tendency for individuals to connect with others who share similar attributes. Exploring this phenomenon is crucial for understanding a wide spectrum of social behaviors, including the spread of misinformation and the dynamics of societal debates. In this study, we leverage a graph transformation strategy-which analyzes the interplay between individuals' personal preferences and their structural connections-to investigate mechanisms of opinion/information diffusion. Among these latter ones, we focus on the Deffuant-Weisbuch model to simulate opinion dynamics and the Independent Cascade model to simulate information spread. Our findings on real-world social networks suggest that emphasizing attribute similarities enhances graph cohesion, whereas forcing structural similarities leads to fragmentation. Moreover, we observe a trend towards consensus opinion formation when enhancing attribute similarities, and faster as well as complete coverage of information spread in the same setup. These results motivate the importance of considering both individual attributes and network structure in studying social dynamics.
Human performance augmentation exoskeleton is beneficial in human motion assistance and enhancement. Human muscle force is an important indicator to evaluate the assistance of an exoskeleton. This study applied an inv...
ISBN:
(纸本)9789819607853;9789819607860
Human performance augmentation exoskeleton is beneficial in human motion assistance and enhancement. Human muscle force is an important indicator to evaluate the assistance of an exoskeleton. This study applied an inverse dynamics method based on muscle synergy to analyze human muscle force of the human-exoskeleton hybrid system. We conducted experiments with three subjects equipped with a hip joint-assisted exoskeleton walking at various speeds. Kinematic data, ground reaction forces and electromyography (EMG) signals were measured, and joint torques and forces of representative lower-limb muscles were calculated. The results indicated that both human hip torque and ankle torque were reduced due to wearing the hip joint-assisted exoskeleton, which implied assistance could be transferred across joints. The results also showed the effects of the exoskeleton on muscle force modulation. The dynamics analysis with muscle force calculation of this study may provide more insights of exoskeleton assistance, and thus be helpful in evaluation of performance augmentation and in design optimization of exoskeleton.
Bionic soft hands have attracted extensive attention due to their high flexibility, safety, and adaptability, enabling them to mimic human hand movements such as grasping. This paper presents the design a pneumatic-dr...
ISBN:
(纸本)9789819607976;9789819607983
Bionic soft hands have attracted extensive attention due to their high flexibility, safety, and adaptability, enabling them to mimic human hand movements such as grasping. This paper presents the design a pneumatic-driven sorting modular soft hand based on visual object detection. Additionally, a comprehensive sorting system control platform was constructed, integrating vision sensor, the modular soft hand, object detection, drive units, and mobile sorting strategies. To verify the system's performance, we established datasets for three types of objects: Lightweight building blocks, fragile eggs, and fresh oranges, and utilized the YOLOv8 algorithm to train the object detection model, thus accurately identifying the type of objects to be sorted and the three-dimensional coordinates of their grasping points. The experimental results demonstrate that this control platform is efficient, stable, and non-destructive in both single-target and multi-target mixed sorting, fully verifying the system's practicality and reliability.
暂无评论