Contrastive Learning (CL) enhances the training of sequential recommendation (SR) models through informative self-supervision signals. Existing methods often rely on data augmentation strategies to create positive samples and promote representation invariance. Some strategies such as item reordering and item substitution may inadvertently alter user intent. Supervised Contrastive Learning (SCL) based methods find an alternative to augmentation-based CL methods by selecting same-target sequences (interaction sequences with the same target item) to form positive samples. However, SCL-based methods suffer from the scarcity of same-target sequences and consequently lack enough signals for contrastive learning. In this work, we propose to use similar sequences (with different target items) as additional positive samples and introduce a Relative Contrastive Learning (RCL) framework for sequential recommendation. RCL comprises a dual-tiered positive sample selection module and a relative contrastive learning module. The former module selects same-target sequences as strong positive samples and selects similar sequences as weak positive samples. The latter module employs a weighted relative contrastive loss, ensuring that each sequence is represented closer to its strong positive samples than its weak positive samples. We apply RCL on two mainstream deep learning-based SR models, and our empirical results reveal that RCL can achieve 4.88% improvement averagely than the state-of-the-art SR methods on five public datasets and one private dataset.

Mental disorders are among the foremost contributors to the global healthcare challenge. Research indicates that timely diagnosis and intervention are vital in treating various mental disorders. However, the early somatization symptoms of certain mental disorders may not be immediately evident, often resulting in their oversight and misdiagnosis. Additionally, the traditional diagnosis methods incur high time and cost. Deep learning methods based on fMRI and EEG have improved the efficiency of the mental disorder detection process. However, the cost of the equipment and trained staff are generally huge. Moreover, most systems are only trained for a specific mental disorder and are not general-purpose. Recently, physiological studies have shown that there are some speech and facial-related symptoms in a few mental disorders (e.g., depression and ADHD). In this paper, we focus on the emotional expression features of mental disorders and introduce a multimodal mental disorder diagnosis system based on audio-visual information input. Our proposed system is based on spatial-temporal attention networks and innovative uses a less computationally intensive pre-train audio recognition network to fine-tune the video recognition module for better results. We also apply the unified system for multiple mental disorders (ADHD and depression) for the first time. The proposed system achieves over 80% accuracy on the real multimodal ADHD dataset and achieves state-of-the-art results on the depression dataset AVEC 2014.

The linear quadratic (LQ) optimal control problem with indefinite weighting matrices and the stabilization problem for discrete-time descriptor stochastic Markov jump systems (DSMJSs) involving state-dependent noises are studied. By using the Moore-Penrose generalized inverse of matrices and the equivalent transformation of restricted system, the indefinite LQ problem for DSMJSs is equivalently converted into the indefinite LQ problem for Markov jump systems (MJSs). Under some viable conditions and stabilization assumption, the generalized stochastic algebraic Riccati equation having a unique semi-positive definite solution is guaranteed. Then the necessary and sufficient conditions which ensure that DSMJSs are causal and mean-square stable are established. It is shown that the admissibility of the optimal closed-loop systems is equivalent to stabilizability in mean-square sense of the transformed MJSs. Besides, an efficient iterative algorithm is given to verify the mean-square stabilizability of DSMJSs by solving an optimization problem. Two examples including a practical RLC circuit system are presented as verifications of the theoretical results.

The robust LQ optimal regulator problem for discrete-time uncertain singular Markov jump systems (SMJSs) is solved by introducing a new quadratic cost function established by the penalty function method, which combines the penalty function and the weighting matrices. First, the indefinite robust optimal regulator problem for uncertain SMJSs is transformed into the robust optimal regulator problem with positive definite weighting matrices for uncertain Markov jump systems (MJSs). The transformed robust LQ problem is settled by the robust least-squares method, and the condition of the existence and analytic form of the robust optimal regulator are proposed. On the infinite horizon, the optimal state feedback is obtained, which can guarantee the regularity, causality, and stochastic stability of the corresponding optimal closed-loop system and eliminate the uncertain parameters of the closed-loop system. A numerical example and a practical example of DC motor are used to verify the validity of the conclusions.

This work investigates the indefinite linear quadratic (LQ) optimal control problem for discrete-time rectangular descriptor Markov jump systems (DMJSs) with additive noise on finite and infinite horizon, where the weight matrices of quadratic cost function are only symmetric. Under a set of equivalent transformations, the indefinite LQ problem for rectangular DMJSs is equivalently turned into indefinite LQ problem for Markov jump systems (MJSs). On finite horizon, sufficient and necessary conditions are given for the solvability of the transformed indefinite LQ problem, and the exact optimal control and the optimal cost value are derived. Then, sufficient and necessary conditions are derived to guarantee that the transformed equivalent LQ problem for MJSs is definite, meanwhile, the unique optimal control and the non-negative optimal cost value are acquired. Besides, on infinite horizon, to ensure that the dynamic part for the optimal closed-loop system has a unique solution and is stochastically stable, several sufficient and necessary conditions are provided. Finally, two examples are presented as verifications of the theoretical results.

Attention Deficit Hyperactivity Disorder (ADHD) causes significant impairment in various domains.Early diagnosis of ADHD and treatment could significantly improve the quality of life and functioning.Recently, machine learning methods have improved the accuracy and efficiency of the ADHD diagnosis process.However, the cost of the equipment and trained staff required by the existing methods are generally huge.Therefore, we introduce the video-based frame-level action recognition network to ADHD diagnosis for the first time.We also record a real multi-modal ADHD dataset and extract three action classes from the video modality for ADHD diagnosis.The whole process data have been reported to CNTW-NHS Foundation Trust, which would be reviewed by medical consultants/professionals and will be made public in due course.

Cross-domain recommendation (CDR) is an effective way to alleviate the data sparsity problem. Content-based CDR is one of the most promising branches since most kinds of products can be described by a piece of text, especially when cold-start users or items have few interactions. However, two vital issues are still under-explored: (1) From the content modeling perspective, sufficient long-text descriptions are usually scarce in a real recommender system, more often the light-weight textual features, such as a few keywords or tags, are more accessible, which is improperly modeled by existing methods. (2) From the CDR perspective, not all inter-domain interests are helpful to infer intra-domain interests. Caused by domain-specific features, there are part of signals benefiting for recommendation in the source domain but harmful for that in the target domain. Therefore, how to distill useful interests is crucial. To tackle the above two problems, we propose a metapath and multi-interest aggregated graph neural network (M2GNN). Specifically, to model the tag-based contents, we construct a heterogeneous information network to hold the semantic relatedness between users, items, and tags in all domains. The metapath schema is predefined according to domain-specific knowledge, with one metapath for one domain. User representations are learned by GNN with a hierarchical aggregation framework, where the intra-metapath aggregation firstly filters out trivial tags and the inter-metapath aggregation further filters out useless interests. Offline experiments and online A/B tests demonstrate that M2GNN achieves significant improvements over the state-of-the-art methods and current industrial recommender system in Dianping, respectively. Further analysis shows that M2GNN offers an interpretable recommendation.

In order to maintain human safety in autonomous vehicles, pedestrian detection and tracking in real-time have become crucial research areas. The critical challenge in this field is to improve pedestrian detection accuracy while reducing tracking processing time. Due to the fact that pedestrians move in groups with the same speed and direction, we can address this challenge by detecting and tracking pedestrian groups. This work focused on pedestrian group detection. Various clustering methods were used in this study to identify pedestrian groups. Firstly, pedestrians were identified using a convolutional neural network approach. Secondly, K-Means and DBSCAN clustering methods were used to identify pedestrian groups based on the coordinates of the pedestrians’ bounding boxes. Moreover, we proposed a modified DBSCAN clustering method named GDSCAN that employs dynamic epsilon to different areas of an image. The experimental results on the MOT17 dataset show that GDSCAN outperformed K-Means and DBSCAN methods based on the Silhouette Coefficient score and Adjusted Rand Index (ARI).

Recently, electrical vehicles (EVs) have attracted considerable attention from researchers due to the transition of the transportation industry and the increasing demand in the clean energy domain. State of charge (SOC) of Li-ion batteries has a significant role in improving the efficiency, performance, and reliability of EVs. Estimating the SOC of the Li-ion battery cannot be done directly from inner measurements due to the complex and dynamic nature of these kinds of batteries. Several data-driven approaches have recently been used to estimate the SOC of Li-ion batteries, benefiting from the availability of battery data and hardware computing capacity. However, selecting the discriminative features and best supervised machine learning (ML) models for accurate battery states estimation is still challenging. Thus, this paper investigates the effect of different ML models and extracted input features of Li-ion batteries, including Electrochemical Impedance Spectroscopy (EIS) and multi-channel feature set on the SOC prediction. The results on the public Panasonic dataset indicate that using EIS feature set as an input to the deep neural network (DNN) model is more efficient than the multi-channel feature set. Moreover, the DNN model outperforms the Gaussian process regression (GPR) model in terms of the mean squared error, mean absolute error, and root mean squared error rates for the SOC prediction.

Reasonable Aurora classification is particularly important for studying the relationship between aurora phenomena and the process of magnetosphere dynamics. With the development of computer science, image processing and pattern recognition technology, new approaches for Aurora classification are springing up. In this paper, we extract the LBP feature of images and use the distributed weighted KNN based on optimal discriminant dictionary for sparse representation as the classification method to discriminate the shape of aurora. The proposed method combines compressed sensing approaches and distributed computing technology, improving the accuracy and effectiveness of the existed sparse representation methods. The experimental results show that the proposed method significantly enhances the power of discrimination of aurora features, and consequently improve the accuracy and effectiveness of the classification of aurora images.