In recent years, the combination of the unmanned aerial vehicle (UAV) and the Internet of Things (IoT) has been applied to various research tasks assisting in collecting data in fieldwork. However, often the mainstream object detection models cannot be directly transformed to deal with drone images. This drawback seriously hinders the development of interdisciplinary applications such as AI for agriculture or forestry. For the microatoll detection for monitoring marine fishery, this study investigates the characteristics of microatolls in drone images. Based on the analysis of the dataset, a novel generative adversarial network-based microatoll detection network is proposed that successfully converts a segmentation model to improve the microatoll detection task. The model significantly elevates the performance of the selected baseline in the microatoll centre detection task (by increasing recall from 25% to 55% and precision from 57% to 64%), and it also removes the use of the extra anchor and post-processing step.

Recommender systems play a crucial role in personalizing user experiences by analyzing vast amounts of user interaction records to suggest relevant items. However, the use of sensitive user information raises significant privacy concerns, particularly in light of regulations such as the General Data Protection Regulation. In response to these challenges, this paper introduces Rapid Collaborative Machine Unlearning (RaCoMU), a novel framework designed to efficiently handle user data deletion requests while preserving recommendation quality. RaCoMU employs a two-step data partitioning strategy that consolidates user data within individual subsets, enhancing collaborative information utilization. Additionally, a collaboration-based attention aggregation method is proposed, which weights contributions of the submodels based on user similarities. Our extensive experiments on real-world datasets across three recommendation models demonstrate that the proposed RaCoMU achieves superior unlearning efficiency and significantly outperforms existing frameworks in model effectiveness. The findings underscore the potential of RaCoMU to balance user privacy and recommendation performance effectively.

Temperature forecasting significantly influences personal lives and manufacturing industry, making research into temperature prediction vital for modern society. However, the existing mainstream methods, such as convolutional neural network (CNN) or recurrent neural network (RNN) based models, struggle with capturing complex non-linear relationships and multi-scale dependencies, limiting their accuracy. To address these issues, we propose a Kolmogorov-Arnold Networks (KAN) based model, KAN-Temporal Spectral ConvNet (KAN-TSCN), which integrates frequency time and eigenspace domain processing to capture intricate spatial and temporal dependencies. Experimental results demonstrate that the proposed KAN-TSCN outperforms the selected baseline models across three real-world datasets in terms of MSE, MAE, and R2metrics.

The explosion of cameras embedded in IoT devices—from mobile phones to autonomous vehicles—has positioned video analytics as a transformative AI tool across healthcare, smart cities, and beyond. Yet, the substantial computing and bandwidth demands of these applications outstrip what IoT devices alone can handle, particularly when low latency is required. Computing Power Networking (CPN) is an emerging solution that unifies cloud, edge, and device resources, enabling seamless, efficient task distribution for real-time analytics. While recent advances in cloud-edge frameworks show promise, current approaches often neglect the economic incentives that drive resource availability. To address this, we present a novel, privacy-enabled dual-layer Stackelberg game model that establishes a dynamic pricing strategy for video analytics in CPN. Our model introduces a two-stage negotiation: IoT devices contract with edge servers for computational and bandwidth resources, while edge servers may offload tasks to the cloud for enhanced service. Using game theory, we derive optimal pricing and offloading strategies under both complete and incomplete information, proving a Nash equilibrium. Comprehensive simulations validate our approach, showing improvements in resource efficiency, reduced latency, and incentivized resource-sharing across all CPN tiers. Specifically, our hybrid offloading strategy significantly reduces latency compared to edge-only and cloud-only computation models. For varying IoT device quantities, the average latency reduction across all scenarios is approximately 30.5%. This work provides an economically sustainable, privacy-conscious solution to the computational challenges of video analytics in an interconnected, resource-sharing ecosystem.

Vehicular edge computing (VEC) tackles the escalating computational demands of intelligent transportation systems by offloading tasks to nearby roadside units (RSUs) for processing. However, in the dynamic vehicular network environment, where vehicles are constantly moving, effective VEC demands a sophisticated approach to managing computing, caching, and communication resources. This involves coordinating resource allocation and data caching across multiple vehicles and RSUs while making complex decisions about task placement. In this paper, we present VECO, a Vehicular Edge Caching and Offloading framework powered by digital twins (DTs). VECO leverages DTs for real-time monitoring of network conditions and resource states, enabling predictive analysis and intelligent decision-making. The framework incorporates a Dynamic Task Caching and Computation Offloading (DT2C) mechanism to optimize data caching and adapt task offloading based on task characteristics and dynamic resource availability. Specifically, we develop a utility-based caching algorithm for RSUs and a novel task offloading strategy using a Proximal Policy Optimization-based deep reinforcement learning algorithm. Extensive experiments demonstrate that VECO, augmented by the DT2C mechanism, significantly outperforms baseline approaches, achieving faster learning convergence and a 21% reduction in total costs, including system latency and energy consumption.

Predicting student performance (PSP) is critical to intelligent tutoring systems in online education services. Accurate predictions enable data-driven decision making and facilitate the implementation of timely educational interventions. While previous approaches, such as cognitive diagnosis models and data mining techniques, have demonstrated effectiveness with small-scale and static datasets, they face significant challenges in large-scale online learning environments. In these settings, vast volumes of learning responses are continuously generated as students engage with exercises. Those responses are characterized by high dimensionality and incompleteness (HDI) and often manifest as streaming data, which limits the applicability of most existing prediction methods, therefore posing a new challenge to traditional PSP tasks. To remedy the void of PSP in the HDI and incremental scenario, we propose an incremental nonlinear co-latent factor analysis (IN-CoLFA) model, which enhances latent factor analysis – an element-wise learning framework – by incorporating a co-factorization technique and integrating a neural network-inspired structure. To facilitate incremental learning, we develop momentum-accelerated stochastic gradient-based algorithms, which enable the model to perform offline training on historical data and continuously refine its predictions as new student performance becomes available. Experiments on several real-world datasets demonstrate the efficacy and efficiency of our approach in both stationary and incremental PSP tasks under HDI conditions.

Within the domain of intelligent education, also known as smart education, the recommender system propelled by deep learning strives to attain exemplary model performance. However, deep learning models invariably involve the processing of voluminous user privacy data during training phases, and they subjugate themselves to substantial risks of privacy breaches concerning both students and educators. Traditional approaches involving retraining the entire dataset and classical privacy protection methods such as differential privacy and homomorphic encryption struggle to balance model performance and training time expenditure. This presents significant challenges for individuals and enterprises in managing privacy concerns. While balancing personal and corporate interests in privacy protection, Machine Unlearning reveals its potential as a productive strategy to navigate these challenges. This study compares the time cost and performance of model retraining using Machine Unlearning with those of retraining using conventional approaches. The experiment results show that the use of Machine Unlearning algorithms not only effectively protects privacy but also significantly reduces the time required for model retraining. Furthermore, the performance of models employing Machine Unlearning is essentially congruent with that of retraining a model with an entire dataset.

Applying knowledge distillation (KD) in federated learning (FL) can transfer model knowledge between clients’ local models and global model, which helps to improve the generalization of the global model. However, this requires both the clients and the server to have public data sets, which may lead to potential privacy disclosure issues. In this paper, we propose a federated data-free knowledge distillation framework, namely FedDFKD, which does not rely on any public data sets. There is a lightweight delivery model we design to learn and transfer model knowledge in different clients. During local training, the local model is jointly trained with delivery model using local data sets, and the local model feeds back its knowledge to the delivery model after it has finished its training phase in this communication round. Afterwards, the server performs global model aggregation and knowledge distillation of the delivery model. Finally, the server returns global model and distillation result to clients. We compare FedDFKD with the most representative aggregation algorithms in FL, and the results show that our method is feasible and outperforms the compared methods by between 0.1 and 3.96 percent of the global model on the MNIST dataset.

As an emerging pedagogy, micro learning aims to make use of people's fragmented spare time and provide personalized online learning service, for example, by pushing fragmented knowledge to specific learners. In the context of big data, the recommender system is the key factor for realizing the online personalization service, which significantly determines what information will be fmally accessed by the target learners. In the education discipline, due to the pedagogical requirements and the domain characteristics, ranking recommended learning materials is essential for maintaining the outcome of the massive learning scenario. However, many widely used recommendation strategies in other domains showed defectiveness in the ability to rank the recommended results. In this paper, we propose a novel recommendation strategy based on the combination of the language model and the translation model. The proposed recommendation strategy aims to filter out unsuitable learning materials and ranks the recommended learning materials more effectively.

Micro learning aims to assist users in making good use of smaller chunks of spare time and provides an effective online learning service. However, to provide such personalized online services on the Web, a number of information overload challenges persist. Effectively and precisely mining and extracting valuable information from massive and redundant information is a significant pre-processing procedure for personalizing online services. In this study, we propose a deep sequence labelling model for locating, extracting, and classifying key information for micro learning services. The proposed model is general and combines the advantages of different types of classical neural network. Early evidence shows that it has satisfactory performance compared to conventional information extraction methods such as conditional random field and bi-directional recurrent neural network, for micro learning services.

With the fast development of Internet technologies and mobile devices, space-time boundaries of people's daily activities become blurred. This makes people utilizing their fragmented time become possible. In recent years, micro learning, which aims to make good use of people's fragmented time and deliver micro-format of open education resource to learners, has drawn wider attention. The generation and delivery of micro learning materials are two essential steps for the micro learning service. This paper first discusses the characteristics of three significant stages of a sophisticated micro learning system: segmentation, annotation, and recommendation, for learning materials. Then various state-of-the-art techniques for different processing stages are reviewed in this survey. Different segmentation and annotation strategies based on different information sources (such as content and users' interaction) are demonstrated and analysed. Soft computing, transfer learning, reinforcement learning, and context-aware techniques are also compared and discussed for solving different difficulties in recommending scenarios. We contribute this paper as the first work focusing on the three-phased techniques in micro learning.

In this paper, we present an online computation approach implemented in a cloud-based system to assist open education resource (OER) providers and instructors dealing with the sparsity of data in micro OER recommendation. An algorithmic framework is provided to realize the novel micro OER recommendation system based on heuristic rules. These rules can also optimize the approaches to blending new-coming micro OERs into established learning paths. Comparing with different widely used recommender systems, our evaluation shows the proposed heuristic algorithms for online computation performs satisfactorily in terms of precision and recall values.

Recent advances in the steel industry have encountered challenges in soliciting decision making solutions for quality control of products based on data mining techniques. In this paper, we present a steel quality control prediction system encompassing with real-world data as well as comprehensive data analysis results. The core process is cautiously designed as a regression problem, which is then best handled by grouping various learning algorithms with their massive resource of historical production datasets. The characteristics of the currently most popular learning models used in regression problem analysis are as well investigated and compared. The performance indicates our steel quality control prediction system based on ensemble machine learning model can offer promising result whilst delivering high usability for local manufacturers to address the production problem by aid of development of machine learning techniques. Furthermore, real-world deployment of this system is demonstrated and discussed. Finally, future directions and the performance expectation are pointed out.