University of La Rochelle

Purpose Although mobile payment (m-payment) adoption has grown rapidly worldwide, existing research remains fragmented in explaining the factors that drive users' continued use beyond initial adoption. Prior reviews have predominantly concentrated on adoption-related factors and acceptance models, with limited attention to post-adoption behaviours such as continuance intention (CI). This study aims to address this gap by conducting a systematic review and proposing a conceptual framework grounded in Shannon and Weaver’s communication model. Design/methodology/approach A systematic review of 99 peer-reviewed journal articles from 2014 to 2024 was conducted following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Articles were retrieved from five academic sources and screened through a four-stage process. Influencing factors were categorized using the adapted communication model into three dimensions: source, channel, and recipient, while also considering mediating, moderating and external factors. Findings The review identifies three core categories of factors influencing m-payment CI: (1) source-related (e.g. satisfaction, trust and confirmation); (2) channel-related (e.g. usefulness, ease of use and security); and (3) recipient-related (e.g. merchant pro-activeness). Mediating variables (e.g. satisfaction and trust), moderating variables (e.g. gender, age and experience) and external influences (e.g. social influence and subjective norms) also play significant roles. Originality/value This study extends communication theory into the digital finance domain by developing a multidimensional source–channel–recipient framework. It applies a classical communication model to synthesize and interpret dispersed continuance research, highlighting underexplored areas such as recipient responsiveness and environmental influences. The framework provides structured implications for researchers, service providers and policymakers seeking to foster sustained m-payment use.

The rapid adoption of electric vehicles (EVs) has heightened the need for advanced diagnostic frameworks to ensure the safety, reliability, and longevity of battery systems. Batteries, as the cornerstone of EV performance, are susceptible to capacity degradation, thermal instability, and anomalies that compromise efficiency and safety. Traditional diagnostic approaches, relying on physics-based models and heuristic thresholds, often fail to address the complexity and variability of real-world EV operations, necessitating scalable, efficient, and privacy-preserving solutions. This paper presents a Split Multi-Task Federated Learning (SMTFL) framework that leverages a Transformer-based architecture to address the dual challenges of battery health anomaly detection and capacity estimation. By combining the efficiency of split learning with the collaborative benefits of federated learning, the proposed system enables EVs to train and share partial model updates while retaining sensitive data locally. Experimental results on a recent large-scale EV dataset demonstrate the superior performance of the SMTFL framework, achieving an accuracy of 98.9% for battery health anomaly detection and a mean absolute error (MAE) of 0.32 Ah for capacity estimation, corresponding to a relative error of approximately 1.80%, significantly outperforming state-of-the-art approaches. By training a single model for both tasks and performing computations during charging sessions, the framework minimizes energy consumption and computational overhead, ensuring practicality for large-scale deployment in real-world EV systems.

Of all the Nobel winners in literature, Pearl S. Buck has probably been treated as an outsider for the longest time.Chinese

Abstract Despite their richness in bioactive compounds, green onion (Allium cepa) leaves are rarely valorized. This study compares several drying methods in terms of product quality and process performance: continuous (CCD), intermittent (ICD), interval starting accessibility (ISAD) while, freeze-drying (FD) was used as a control treatment. To achieve the same final water activity, the effective ISAD and ICD drying times were 10 and twice times lower than CCD. During drying, the core temperature of the onion leaves remained stable at 22 °C for ISAD while it reached the drying air temperature during (45 °C) during CCD and 34 °C for ICD. ICD and ISAD dried leaves had a total color difference between a maximum for CCD and a minimum for FD leaves. Moreover, CCD and ICD onion leaves had the highest loss of total phenolic content, while FD and ISAD achieved the highest preservation.

International audience

Blockchain-based Federated Learning (BFL) is an emerging decentralized machine learning paradigm that enables model training without relying on a central server. Although some BFL frameworks are considered privacy-preserving, they are still vulnerable to various attacks, including inference and model poisoning. Additionally, most of these solutions employ strong trust assumptions among all participating entities or introduce incentive mechanisms to encourage collaboration, making them susceptible to multiple security flaws. This work presents VerifBFL, a trustless, privacy-preserving, and verifiable federated learning framework that integrates blockchain technology and cryptographic protocols. By employing zero-knowledge Succinct Non-Interactive Argument of Knowledge (zk-SNARKs) and in-crementally verifiable computation (IVC), VerifBFL ensures the verifiability of both local training and aggregation processes. The proofs of training accuracy and aggregation are verified on-chain, guaranteeing the integrity and auditability of each participant's contributions. To protect training data from inference attacks, VerifBFL leverages differential privacy. Finally, to demonstrate the efficiency of the proposed protocols, we built a proof of concept using emerging tools. The results show that generating proofs for local training and aggregation in VerifBFL takes less than 81s and 2s, respectively, while verifying them on-chain takes less than 0.6s.

Edge Computing (EC) extends computing functionalities from remote cloud data centers to the proximity of end-user devices at the network edges, thereby reducing application response time. However, existing solutions face challenges in scalability, efficient resource management, and near real-time adaptation in dynamic environments. In this paper, we jointly investigate the application placement and load distribution challenges in an EC-enabled mobile network. We propose a hierarchical distributed Limited Look-Ahead Control (LLC) approach that mitigates centralized control bottlenecks by breaking down the overall decision problem into a series of local problems solved cooperatively through a two-tier architecture. The global controller processes system-wide information and sets local constraints, while local controllers make autonomous decisions coordinated through mutual information exchange. Utilizing LLC allows for anticipation and adaptation to load fluctuations. Even though the results indicate that the trade-off between system performance and scalable decisions depends on how the overall problem is decomposed, our distributed solution significantly reduces the controller’s execution time compared to a centralized approach.

Bycatch (e.g., the accidental capture of non-targeted species by fisheries) is a leading cause of human-induced mortality, contributing to significant populations declines worldwide. Often stemming from the overlap between food resources and fishery t

Accurate classification of coffee leaf diseases is essential for early intervention and improved crop yield. However, training deep learning models for this task often requires large labeled datasets, which can be expensive and time-consuming to obtain. To address this challenge, we propose an iterative active learning framework with diverse expansion sampling for coffee leaf disease classification, aimed at the need to reduce annotation costs while maintaining high classification accuracy in agricultural disease detection. Our method iteratively selects the most informative and representative samples for labeling, reducing annotation costs while maintaining high classification performance. Diverse expansion sampling leverages a per-class diversity-based selection method inspired by Greedy Furthest-Point Selection, to effectively capture the data distribution. Experimental results demonstrate that our approach significantly improves classification accuracy with fewer labeled samples compared to traditional supervised learning. The proposed method is evaluated on two coffee leaf disease datasets from different countries across various data scenarios. The results indicate that our active learning strategy achieves an improvement of 1.7% and 0.9% in F1 score on VIECOLD and BRACOL, respectively, using the same amount of labeled data as in the supervised approach. The results also show that using only 35% of labeled data for VIECOLD and 50% for BRACOL surpasses the performance achieved with the full dataset when applied to the same baseline model. This emphasizes the advantages of selective data utilization for enhanced generalization and efficiency. This research highlights the potential of active learning in optimizing data annotation for agricultural applications, making deep learning-based plant disease detection more accessible and efficient.