Applied Intelligence最新文献

The main contribution is the design, implementation and validation of a complete AI-driven electronic nose architecture to perform the classification of whiskey and acetones. This classification is of paramount important in the distillery production line of whiskey in order to predict the quality of the final product. In this work, we investigate the application of an e-nose (based on arrays of single-walled carbon nanotubes) to the distinction of two different substances, such as whiskey and acetone (as a subproduct of the distillation process), and discrimination of three different types of the same substance, such as three types of whiskies. We investigated different strategies to classify the odor data and provided a suitable approach based on random forest with accuracy of 99% and with inference times under 1.8 seconds. In the case of clearly different substances, as subproducts of the whiskey distillation process, the procedure presented achieves a high accuracy in the classification process, with an accuracy around 96%.

Graph neural networks (GNNs) demonstrate their effectiveness in facilitating node classification and a range of graph-based tasks. However, recent studies have revealed that GNNs can be vulnerable to various adversarial attacks. Despite various defense strategies, ranging from attack-agnostic defenses to attack-oriented defenses that have been proposed to mitigate the impact of adversarial attacks on graph data, effectively learning attack-agnostic graph representation remains an open challenge. This paper introduces a novel information Competition-based framework for Graph Neural Networks (i.e., iC-GNN, e.g., iC-GCN, iC-GAT, etc.) to enhance the robustness of GNNs against various adversarial attacks in node classifications. Through the use of graph reconstruction and low-rank approximation, our approach learns diversified graph representations to collaboratively perform node classifications. Meanwhile, mutual information constraints are utilized on different graph representations to ensure diversity and competition in graph features. The experimental results indicate that within the proposed framework, iC-GCN outperforms other graph defense frameworks in countering a wide range of targeted and non-targeted adversarial attacks in both evasion and poisoning training scenarios. Additionally, this concept has been extended to encompass other widely utilized GNN models like iC-GAT and iC-SAGE. All iC-GNN models demonstrate effective defense capabilities, demonstrating comparable resilience to adversarial attacks. This underscores the superiority and scalable nature of the iC-GNN framework, providing opportunities for a variety of graph learning applications.

Greenhouses are a critical component of modern agriculture, facilitating crop growth and development, and accurate predictions of temperature and humidity are essential for mitigating crop diseases and optimizing the growth environment. However, short- and medium-term forecasts of temperature and humidity are challenging because of the complexity of greenhouse microclimates. This paper presents a hybrid model that integrates a frequency-enhanced channel attention mechanism optimized with a temporal convolutional network (TCN-FECAM) and an iTransformer. The model employs a cross-attention mechanism incorporating the advantages of the two models, and a 48-sequence sliding window strategy is used to ensure accurate multistep predictions of temperature and humidity over spans of 3 h to 24 h. The experimental results demonstrate that the TCN-FECAM-iTransformer model outperforms other models across diverse time scales, including GRU, LSTM, Informer, Autoformer, Crossformer, FAM-LSTM, and TPA-LSTM. Specifically, in temperature prediction, the model achieves R² coefficients of 0.979, 0.973, 0.968, and 0.953 and RMSE values of 0.657, 0.806, 0.923, and 1.126, for 3 h, 6 h, 12 h, and 24 h intervals, respectively. In humidity prediction, the model obtains R² coefficients of 0.976, 0.961, 0.947, and 0.939 and RMSE values of 1.805, 2.567, 3.132, and 3.451 for 3 h, 6 h, 12 h, and 24 h intervals, respectively. The model therefore exhibits reliable performance in predicting temperature and humidity in greenhouse environments, offering robust support for monitoring and early warnings in crop growth environments.

This paper proposes a novel method for network latency estimation. Network latency estimation is a crucial indicator for evaluating network performance, yet accurate estimation of large-scale network latency requires substantial computation time. Therefore, this paper introduces a method capable of enhancing the speed of network latency estimation. The paper represents the data structure of network nodes as matrices and introduces a time dimension to form a tensor model, thereby transforming the entire network latency estimation problem into a tensor completion problem. The main contributions of this paper include: optimizing leveraged sampling for tensors to improve sampling speed, and on this basis, introducing the Qatar Riyal (QR) decomposition of tensors into the Alternating Direction Method of Multipliers (ADMM) framework to accelerate tensor completion; these two components are combined to form a new model called LNLS-TQR. Numerical experimental results demonstrate that this model significantly improves computation speed while maintaining high accuracy.

Achieving consensus tracking control of a multiagent system (MAS) is challenging. This article proposes an innovative consensus control scheme of a MAS that is composed of electromechanical actuators. The open-loop derivative-type iterative learning control (ILC) is adopted as the baseline consensus controller. The baseline controller has systematically evolved to a proportional-derivative-type ILC to achieve better consensus tracking control for the said actuator. The proposed ILC procedure is synthesized by including the weighted sum of the tracking error as well as the tracking error-derivative variables. The respective learning gains of the aforementioned tracking error variables are pre-calibrated to ensure faster trajectory tracking with better accuracy. The PD-type ILC law strengthens the system’s disturbance resilience and improves its asymptotic convergence rate. The designed controllers are tested on two different communication topologies via simulations and reliable hardware experiments, in which the virtual leader provides the desired trajectory to four agents. Only the fixed agents interact with the leader to obtain the desired trajectory information in different communication topologies. The fixed agent guarantees accurate trajectory tracking behavior by modifying the control effort according to the deviation between its actual trajectory and the trajectories of the neighboring agents and the virtual leader. The corresponding test results indicate that the proposed PD-type ILC significantly enhances the tracking accuracy and the convergence rate of the system compared to the D-type ILC, validating the effectiveness of the proposed control scheme under different communication topologies.

Archipelagic areas are in urgent need of efficient logistics systems to replace the limited bridges and fixed liners. This paper addresses the dynamic pickup and delivery challenge in inter-island logistics by utilizing waterborne autonomous guided vessels (AGVs). Specifically, the waterborne inter-island logistics problem is precisely expressed as a mixed integer programming (MIP) model. The model considers a variety of practical system constraints that could arise for the waterborne AGVs, e.g., capacity, time windows, berth allocation, and loading constraints. Moreover, in order to solve the possible large-scale scheduling problem dynamically and efficiently, we design an improved variable neighborhood search heuristic method. The approach is featured with four local search strategies and an effective perturbation heuristic to deal with the local minima issue. Extensive comparison experiments are carried out using real-world datasets. The results demonstrate that our algorithm outperforms baseline algorithms in 98% of cases, achieving improvements of over 10% compared to greedy rule-based methods and more than 5% over state-of-the-art heuristic algorithms, such as VNSME. These findings highlight the substantial benefits of the proposed technique, offering significant cost savings when effectively implemented. Comprehensive ablation experiments and parameter sensitivity analyses also demonstrate that the proposed algorithm has superior capabilities in space exploration and exploitation, provided that the step size operator is properly set. The proposed modeling and solution algorithms show great potential in enhancing the efficiency of the inter-island logistics systems.

Time series anomaly detection is crucial for Internet of Things (IoT) management and system security. Mainstream methods typically treat time series as an indivisible whole to capture insights into normal patterns. However, time series contain deterministic components such as seasonality, periodicity, and trends. Non-decomposition-based methods often average out these components during training, leading to the over-smoothing of normal time series features. This deviates from the core principle of unsupervised time series anomaly detection: to establish a clear boundary between normal and anomalous patterns. We propose DifNet, a comprehensive unsupervised time series anomaly detection method that effectively mitigates feature over-smoothing (FOS). DifNet adopts the concept of time series decomposition and utilizes Fast Fourier Transform (FFT) to analyze the periodic components of time series, guiding the decomposition and fusion process. Additionally, we design a difference-based multi-resolution decomposition network to thoroughly extract complex periodic dependencies within the time series. For multivariate time series data, DifNet follows the channel independence principle and disregards inter-channel dependencies that introduce redundant information in the data. As a more lightweight alternative, we introduce single-channel autoencoder and cross-channel periodicity adjustment. Meanwhile, DifNet integrates contrastive learning at a fine-grained level to prevent FOS and facilitate the extraction of more distinguishable representations. Extensive experimentation conducted on six multivariate and two univariate time series datasets validates the efficacy of DifNet in time series anomaly detection, DifNet achieved an average improvement in the best F1-score of 14.67% across eight datasets.

In federated learning, performing knowledge distillation on unlabeled proxy data is an effective way to aggregate local models into a global model. Most distillation-based methods assume that all knowledge from local models is contributory, and thereby indiscriminately transfer it to the global model. However, this assumption does not hold in data heterogeneity scenarios. Incorporating noisy knowledge during the distillation can negatively impact the performance of the global model. While filtering the knowledge be transferred is an intuitive solution, performing such filtering in federated learning is challenging due to the lack of available proxy-sample labels for knowledge validation. To address this issue, we propose a knowledge filtering approach for adaptive local model aggregation (FedFAA), which filters the knowledge before distillation based on its relevance. Specifically, we design a scoring method that exploits the representation space of a model to measure the relevance between the model knowledge and each proxy sample, without relying on validation labels. With these relevance scores, we further introduce an adaptive teacher model selection scheme that maintains an appropriate distribution of knowledge-providing teacher models across proxy samples, balancing the precision and diversity of the transferred knowledge after filtering. Theoretical analysis and extensive experiments demonstrate the effectiveness of our approach and its superior performance over six state-of-the-art methods.

Session-based recommendation (SR) aims to predict the next most likely interaction item based on the current sequence of anonymous behaviors. How to learn short- and long-term user preferences is the key to SR research. However, current research ignores the impact of contextual information on users’ short- and long-term preferences when obtaining user preferences. Herein, we propose a Dual-Channel Context-aware Contrastive Learning Graph Neural Networks (DCC-GNN) model for SR. DCC-GNN constructs a time-aware session graph representation learning channel, modeling sessions with temporal context information to learn users’ short-term preferences. To better capture users’ long-term preferences, it also constructs a position correction global graph representation learning channel and uses global session information to learn users’ long-term preferences. To address the issue of data sparsity, contrastive learning techniques are employed to both channels for data augmentation. Finally, a linear combination of the dual-channel session representations serves as the user’s ultimate preference for accurate recommendations. Herein, we performed extensive experiments on three real-world datasets. Experimental results reveal that the performance of the proposed DCC-GNN model demonstrates a considerable improvement compared to baseline models.