Multi-agent reinforcement learning (MARL) has demonstrated significant potential in addressing decision-making challenges. However, in the field of beyond-visual-range (BVR) air combat scenarios, the development of Unmanned Combat Aerial Vehicles (UCAVs) presents new challenges for developing decision-making models, including issues like reward sparsity and non-transitivity. This study presents a novel two-stage training framework tailored for symmetric zero-sum team games, with a focus on 2-v-2 BVR air combat scenarios. Firstly, a lightweight simulation environment for BVR air combat is established with a specially designed observation space and action space. A reward function based on a scripted policy is introduced to improve the efficiency of training. Secondly, a two-stage training framework is employed to train the agents. By incorporating multi-agent reinforcement learning with fictitious self-play, this novel approach enhances policy adaptability against various types of opponents. Finally, the performance of the trained agents is evaluated in the simulated environment. Simulations demonstrate that this innovative approach offers effective and adaptive solutions for autonomous team-based air combat, highlighting its potential to advance the capabilities of UCAVs.

Vision-language models (VLMs) can effectively transfer to various vision tasks via prompt learning. Real-world scenarios often require adapting a model to multiple similar yet distinct tasks. Existing methods focus on learning a specific prompt for each task, limiting the ability to exploit potentially shared information from other tasks. Naively training a task-shared prompt using a combination of all tasks ignores fine-grained task correlations. Significant discrepancies across tasks could cause negative transferring. Considering this, we present Hierarchical Prompt (HiPro) learning, a simple and effective method for jointly adapting a pre-trained VLM to multiple downstream tasks. Our method quantifies inter-task affinity and subsequently constructs a hierarchical task tree. Task-shared prompts learned by internal nodes explore the information within the corresponding task group, while task-individual prompts learned by leaf nodes obtain fine-grained information targeted at each task. The combination of hierarchical prompts provides high-quality content of different granularity. We evaluate HiPro on four multi-task learning datasets. The results demonstrate the effectiveness of our method.

Length of Stay (LoS) estimation is important for efficient healthcare resource management. Since the distribution of LoS is highly skewed, some previous works frame the LoS estimation as a multi-class classification problem by dividing the range of LoS into buckets. However, they ignore the ordinal relationship between labels. The distribution of bucketed LoS, with a heavy head and a heavy tail, is still imbalanced since the long tail is grouped into the last bucket. This paper proposes a Deep Ordinal neural network for Length of stay Estimation in the intensive care units (DOSE). DOSE can exploit the ordinal relationship and mitigate the skewness. The ordinal classification problem is decomposed into a series of binary classification sub-problems by using multiple binary classifiers. To maintain consistency among binary classifiers, the monotonicity constraint penalty is proposed. The number of samples whose labels are higher or lower than a given threshold is at the same level due to the heavy head and tail of the distribution. Therefore, the training data of each binary classifier are balanced. Experiments are conducted on the real-world healthcare dataset. DOSE outperforms all baseline methods in all metrics. The distribution of the prediction of DOSE is more aligned with the ground truth.

In the real world, the class of a time series is usually labeled at the final time, but many applications require to classify time series at every time point. e.g. the outcome of a critical patient is only determined at the end, but he should be diagnosed at all times for timely treatment. Thus, we propose a new concept: Continuous Classification of Time Series (CCTS). It requires the model to learn data in different time stages. But the time series evolves dynamically, leading to different data distributions. When a model learns multi-distribution, it always forgets or overfits. We suggest that meaningful learning scheduling is potential due to an interesting observation: Measured by confidence, the process of model learning multiple distributions is similar to the process of human learning multiple knowledge. Thus, we propose a novel Confidence-guided method for CCTS (C3TS). It can imitate the alternating human confidence described by the Dunning-Kruger Effect. We define the objective-confidence to arrange data, and the self-confidence to control the learning duration. Experiments on four real-world datasets show that C3TS is more accurate than all baselines for CCTS.

Hypergraphs are natural and expressive modeling tools to encode high-order relationships among entities. Several variations of Hypergraph Neural Networks (HGNNs) are proposed to learn the node representations and complex relationships in the hypergraphs. Most current approaches assume that the input hypergraph structure accurately depicts the relations in the hypergraphs. However, the input hypergraph structure inevitably contains noise, task-irrelevant information, or false-negative connections. Treating the input hypergraph structure as ground-truth information unavoidably leads to sub-optimal performance. In this paper, we propose a Hypergraph Structure Learning (HSL) framework, which optimizes the hypergraph structure and the HGNNs simultaneously in an end-to-end way. HSL learns an informative and concise hypergraph structure that is optimized for downstream tasks. To efficiently learn the hypergraph structure, HSL adopts a two-stage sampling process: hyperedge sampling for pruning redundant hyperedges and incident node sampling for pruning irrelevant incident nodes and discovering potential implicit connections. The consistency between the optimized structure and the original structure is maintained by the intra-hyperedge contrastive learning module. The sampling processes are jointly optimized with HGNNs towards the objective of the downstream tasks. Experiments conducted on 7 datasets show shat HSL outperforms the state-of-the-art baselines while adaptively sparsifying hypergraph structures.

Electronic Health Records (EHR) is the repository of patients' involved medical codes in the hospital, including diagnosis codes, medication codes, procedure codes, lab codes, and so on. EHR inherently contains various kinds of relationships such as the code-code, the patient-patient, and the patient-code relationship. Recent research shows that graph representation learning can be an effective tool for capturing complex relationships. However, none of the existing methods considered high-order interactions between patients and medical codes or considered the three relationships together. In this paper, we propose Hypergraph Contrastive Learning (HCL), to jointly learn patient embeddings and code embeddings from the combination of the above three relationships. HCL first constructs a hypergraph from the EHR data. Then, the medical code graph and the patient graph are constructed based on the hypergraph. Empowered with hypergraph attention network, Transformer, and graph attention network, HCL learns representations from three graphs respectively. Next, contrastive learning is applied to aggregate information from these graphs. Finally, the learned representations can support downstream tasks in supervised learning settings and self-supervised learning settings. Experiments are conducted on eICU and MIMIC-III datasets with mortality prediction and readmission prediction tasks. Results show that our method outperforms almost all compared methods on all evaluation metrics and HCL can learn patient representations from medical codes even without labeled data.

The authorization mechanism of smart devices is mainly implemented by firmware, yet many smart devices have security issues about their firmware. Limited research has focused on securing the firmware of smart devices, although increasingly more smart devices are used to deal with the very sensitive applications, activities, and data of users. Thus, research on smart device firmware security is of growing importance. Disassembly is a common method for evaluating the security of authorization mechanisms. When disassembling firmware, the processor type of the running environment and the image base of the firmware should first be determined. In general, the processor type can be obtained by tearing down the device or consulting the product manual. However, it is not easy to determine the image base of firmware. Since the processors of many smart devices are ARM architectures, in this paper, we focus on firmware under the ARM architecture and propose an automated method for determining the image base. By studying the storage law of the jump table in the firmware of ARM-based smart devices, we propose an algorithm, named determining the image base by searching jump tables (DBJT), to determine the image base. The experimental results indicate that the proposed method can successfully determine the image base of firmware, which stores the absolute address in the jump table.

Unsupervised cross-database facial expression recognition (FER) is a challenging problem, in which the training and testing samples belong to different facial expression databases. For this reason, the training (source) and testing (target) facial expression samples would have different feature distributions and hence the performance of lots of existing FER methods may decrease. To solve this problem, in this paper we propose a novel super wide regression network (SWiRN) model, which serves as the regression parameter to bridge the original feature space and the label space and herein in each layer the maximum mean discrepancy (MMD) criterion is used to enforce the source and target facial expression samples to share the same or similar feature distributions. Consequently, the learned SWiRN is able to predict the expression categories of the target samples although we have no access to any label information of target samples. We conduct extensive cross-database FER experiments on CK+, eNTERFACE, and Oulu-CASIA VIS facial expression databases to evaluate the proposed SWiRN. Experimental results show that our SWiRN model achieves more promising performance than recent proposed cross-database emotion recognition methods.

In this paper, we investigate an interesting problem, i.e., unsupervised cross-corpus speech emotion recognition (SER), in which the training and testing speech signals come from two different speech emotion corpora. Meanwhile, the training speech signals are labeled, while the label information of the testing speech signals is entirely unknown. Due to this setting, the training (source) and testing (target) speech signals may have different feature distributions and therefore lots of existing SER methods would not work. To deal with this problem, we propose a domain-adaptive subspace learning (DoSL) method for learning a projection matrix with which we can transform the source and target speech signals from the original feature space to the label space. The transformed source and target speech signals in the label space would have similar feature distributions. Consequently, the classifier learned on the labeled source speech signals can effectively predict the emotional states of the unlabeled target speech signals. To evaluate the performance of the proposed DoSL method, we carry out extensive cross-corpus SER experiments on three speech emotion corpora including EmoDB, eNTERFACE, and AFEW 4.0. Compared with recent state-of-the-art cross-corpus SER methods, the proposed DoSL can achieve more satisfactory overall results.

In this paper, we investigate the performance of artificial noise assisted secure communication in the presence of a full-duplex active eavesdropper who can simultaneously perform eavesdropping and jamming. An approximate closed-form expression for the secrecy rate is derived. With this result, we obtain a new result for the optimal power allocation factor maximizing the secrecy rate, and analyze the impact of self-interference coefficient and jamming power on it. We further consider a more practical scenario where the legitimate user aims to maintain a given target data rate and uses all remaining power for artificial noise (AN) to interfere with the eavesdropper. While the eavesdropper tries to compel the transmitter to reduce the AN by sending jamming signals to the legitimate receiver. To solve this conflict, we introduce a power cost parameter to describe the impact of jamming on the eavesdropper itself and then formulate a game-theoretic framework. We also derive the closed-form equilibrium solutions for both sides. These results show the optimal jamming strategy of the eavesdropper and provide insights into the low bound secrecy performance. Finally, simulation results are provided to verify our analytical results.

The financing of supply chains involves decisions by supply chain members as well as by lending institutions. The optimality of lending decisions in this environment depends on the loss aversion on the part of supply chain members as well as the availability of capital. The purpose of this paper is to understand the impact of capital constraint and loss aversion on operational decisions in supply chains. Traditional models have the bank external to the supply chain, with the bank's interest rate exogenous. This research concerns a capital-constrained supply chain with the manufacturer selling to a loss averse newsvendor-like retailer, and a bank financing both the manufacturer and the retailer. The existence of supply chain finance equilibrium is proven by the use of Stackelberg game analysis. The best pricing and ordering decisions of both manufacturer and retailer are determined, and results demonstrate how these key decisions are influenced by their initial capital and the bank's financial decisions. For instance, the optimal order quantity increases or decreases with initial capital, and it is interesting that bankruptcy protection encourages a cash-constrained retailer to adopt an aggressive ordering strategy. Moreover, it is shown that the retailer's loss aversion has a significant impact on the capital constraint problem. With an increase in loss aversion, the required initial working capital decreases. Loss aversion can even change the retailer's situation from one of capital constraint to one of capital sufficiency. An extension with double orders is given for comparison. Numerical examples are given to demonstrate the impact of initial capital and loss aversion on the optimal decisions and some other managerial insights are discussed.

Compressive Tracking (CT) model is a recently proposed method for visual tracking, in which the appearance model is constructed from the features selected from the multiscale image feature space based on compressive sensing. The CT tracker has been proven to be effective. However, since it does not discriminatively consider the sample importance in its learning procedure, the CT tracker may detect the less important positive samples and, therefore, suffer from drift. In this paper, we present a novel Weighted Compressive Tracking (WCT) model based on the CT tracker. The proposed WCT tracker integrates the sample importance into an efficient online learning procedure so that the features are much more discriminative. Experimental results on challenging benchmark image sequences demonstrate that the proposed WCT tracker performs more favorably than the CT tracker. In addition, the WCT and CT trackers are also applied to the video acquired by the fisheye lens, the result of WCT tracker is very promising, whereas the CT tracker fails.