Complex & Intelligent Systems最新文献

Survival risk prediction models have become important tools for clinicians to improve cancer treatment decisions. In the medical field, using gene expression data to build deep survival neural network models significantly improves accurate survival prognosis. However, it still poses a challenge in building an efficient method to improve the accuracy of cancer-specific survival risk prediction, such as data noise problem. In order to solve the above problem, we propose a diversity reweighted deep survival neural network method with grid optimization (DRGONet) to improve the accuracy of cancer-specific survival risk prediction. Specifically, reweighting can be employed to adjust the weights assigned to each data point in the dataset based on their importance or relevance, thereby mitigating the impact of noisy or irrelevant data and improving model performance. Incorporating diversity into the goal of multiple learning models can help minimize bias and improve learning outcomes. Furthermore, hyperparameters can be optimized with grid optimization. Experimental results have demonstrated that our proposed approach has significant advantages (improved about 5%) in real-world medical scenarios, outperforming state-of-the-art comparison methods by a large margin. Our study highlights the significance of using DRGONet to overcome the limitations of building accurate survival prediction models. By implementing our technique in cancer research, we hope to reduce the suffering experienced by cancer patients and improve the effectiveness of treatment.

Influence Maximization (IM) stands as a central challenge within the domain of complex network analysis, with the primary objective of identifying an optimal seed set of a predetermined size that maximizes the reach of influence propagation. Over time, numerous methodologies have been proposed to address the IM problem. However, one certain network referred to as Imbalanced Heterogeneous Networks (IHN), which widely used in social situation, urban and rural areas, and merchandising, presents challenges in achieving high-quality solutions. In this work, we introduce the Lightweight Reinforcement Learning algorithm with Prior knowledge (LRLP), which leverages the Struc2Vec graph embedding technique that captures the structural similarity of nodes to generate vector representations for nodes within the network. In details, LRLP incorporates prior knowledge based on a group of centralities, into the initial experience pool, which accelerates the reinforcement learning training for better solutions. Additionally, the node embedding vectors are input into a Deep Q Network (DQN) to commence the lightweight training process. Experimental evaluations conducted on synthetic and real networks showcase the effectiveness of the LRLP algorithm. Notably, the improvement seems to be more pronounced when the the scale of the network is larger. We also analyze the effect of different graph embedding algorithms and prior knowledge on algorithmic results. Moreover, we conduct an analysis about some parameters, such as number of seed set selections T, embedding dimension d and network update frequency C. It is significant that the reduction of number of seed set selections T not only keeps the quality of solutions, but lowers the algorithm’s computational cost.

Object detection using UAV images is a current research focus in the field of computer vision, with frequent advancements in recent years. However, many methods are ineffective for challenging UAV images that feature uneven object scales, sparse spatial distribution, and dense occlusions. We propose a new algorithm for detecting small objects in UAV images, called ATBHC-YOLO. Firstly, the MS-CET module has been introduced to enhance the model’s focus on global sparse features in the spatial distribution of small objects. Secondly, the BHC-FB module is proposed to address the large-scale variance of small objects and enhance the perception of local features. Finally, a more appropriate loss function, WIoU, is used to penalise the quality variance of small object samples and further enhance the model’s detection accuracy. Comparison experiments on the DIOR and VEDAI datasets validate the effectiveness and robustness of the improved method. By conducting experiments on the publicly available UAV benchmark dataset Visdrone, ATBHC-YOLO outperforms the state-of-the-art method(YOLOv7) by 3.5%.

Abstract

Multimodal transportation is a modern way of cargo transportation. With the increasing demand for cargo transportation, higher requirements are being placed on multimodal transportation multi-objective routing optimization. In multimodal transportation multi-objective routing optimization, in response to the limitations of classical algorithms in solving large-scale problems with multiple nodes and modes of transport, the limitations of directed transportation networks in the application, and the uncertainty of transport time, this paper proposes an optimization framework based on multi-objective weighted sum Q-learning, combined with the proposed undirected multiple-node network, and characterizes the uncertainty of time with a positively skewed distribution. The undirected multiple-node transportation network can better simulate cargo transportation and characterize transfer information, facilitate the modification of origin and destination, and avoid suboptimal solutions due to the manual setting of wrong route directions. The network is combined with weighted sum Q-learning to solve multimodal transportation multi-objective routing optimization problems faster and better. When modeling the uncertainty of transport time, a positively skewed distribution is used. The three objectives of transport cost, carbon emission cost, and transport time were studied and compared with PSO, GA, AFO, NSGA-II, and MOPSO. The experimental results show that compared with PSO, GA, and AFO using a directed transportation network, the proposed method has a significant improvement in optimization results and running time, and the running time is shortened by 26 times. The proposed method can better solve the boundary of the Pareto front and dominate the partial solutions of NSGA-II and MOPSO. The effect of time uncertainty on the performance of the algorithm is more significant in transport orders with high time weight. With the increase in uncertainty, the reliability of the route decreases. The effectiveness of the proposed method is verified.

Recommender system always suffers from various recommendation biases, seriously hindering its development. In this light, a series of debias methods have been proposed in the recommender system, especially for two most common biases, i.e., popularity bias and amplified subjective bias. However, existing debias methods usually concentrate on correcting a single bias. Such single-functionality debiases neglect the bias-coupling issue in which the recommended items are collectively attributed to multiple biases. Besides, previous work cannot tackle the lacking supervised signals brought by sparse data, yet which has become a commonplace in the recommender system. In this work, we introduce a disentangled debias variational auto-encoder framework (DB-VAE) to address the single-functionality issue as well as a counterfactual data enhancement method to mitigate the adverse effect due to the data sparsity. In specific, DB-VAE first extracts two types of extreme items only affected by a single bias based on the collier theory, which are, respectively, employed to learn the latent representation of corresponding biases, thereby realizing the bias decoupling. In this way, the exact unbiased user representation can be learned by these decoupled bias representations. Furthermore, the data generation module employs Pearl’s framework to produce massive counterfactual data to help fully train the model, making up the lacking supervised signals due to the sparse data. Extensive experiments on three real-world data sets demonstrate the effectiveness of our proposed model. Specifically, our model outperforms the best baseline by 19.5% in terms of Recall@20 and 9.5% in terms of NDCG@100 in the best scenario. Besides, the counterfactual data can further improve DB-VAE, especially on the data set with low sparsity.

Lower limb augmentation exoskeletons (LLAE) have been applied in several domains to enforce human walking capability. As humans can adjust their joint moments and generate different amounts of mechanical energy while walking on different terrains, the LLAEs should provide adaptive augmented torques to the wearer in multi-terrain environments, which requires LLAEs to implement accurate terrain parameter recognition. However, the outputs of previous terrain parameter recognition algorithms are more redundant, and the algorithms have higher computational complexity and are susceptible to external interference. Therefore, to resolve the above issues, this paper proposed a neural network regression (NNR)-based algorithm for terrain slope parameter recognition. In particular, this paper defined for the first time a unified representation of terrain parameters: terrain slope (TS), a single parameter that can provide enough information for exoskeleton control. In addition, our proposed NNR model uses only basic human parameters and LLAE joint motion posture measured by an Inertial Measurement Unit (IMU) as inputs to predict the TS, which is computationally simpler and less susceptible to interference. The model was evaluated using K-fold cross-validation and the results showed that the model had an average error of only 2.09(^circ ). To further validate the effectiveness of the proposed algorithm, it was verified on a homemade LLAE and the experimental results showed that the proposed TS parameter recognition algorithm only produces an average error of 3.73(^circ ) in multi-terrain environments. The defined terrain parameters can meet the control requirements of LLAE in urban multi-terrain environments. The proposed TS parameter recognition algorithm could facilitate the optimization of the adaptive gait control of the exoskeleton system and improve user experience, energy efficiency, and overall comfort.

This paper introduces a novel approach aimed at enhancing the control performance of a specific class of unknown multiple-input and multiple-output nonlinear systems. The proposed method involves the utilization of a fractional-order fuzzy sliding mode controller, which is implemented through online fractional-order reinforcement learning (FOFSMC-FRL). First, the proposed approach leverages two Takagi–Sugeno–Kang (TSK) fuzzy neural network actors. These actors approximate both the equivalent and switch control parts of the sliding mode control. Additionally, a critic TSK fuzzy neural network is employed to approximate the value function of the reinforcement learning process. Second, the FOFSMC-FRL parameters undergo online adaptation using an innovative fractional-order Levenberg–Marquardt learning method. This adaptive mechanism allows the controller to continuously update its parameters based on the system’s behavior, optimizing its control strategy accordingly. Third, the stability and convergence of the proposed approach are rigorously examined using Lyapunov theorem. Notably, the proposed structure offers several key advantages as it does not depend on knowledge of the system dynamics, uncertainty bounds, or disturbance characteristics. Moreover, the chattering phenomenon, often associated with sliding mode control, is effectively eliminated without compromising the system’s robustness. Finally, a comparative simulation study is conducted to demonstrate the feasibility and superiority of the proposed method over other control methods. Through this comparison, the effectiveness and performance advantages of the approach are validated.

The disaster caused by landslide is huge. To prevent the spread of the disaster to the maximum extent, it is particularly important to carry out landslide disaster treatment work. The selection of landslide disaster treatment alternative is a large scale group decision-making (LSGDM) problem. Because of the wide application of social media, a large number of experts and the public can participate in decision-making process, which is conducive to improving the efficiency and correctness of decision-making. A IF-TW-LSGDM method based on three-way decision (TWD) and intuitionistic fuzzy set (IFS) is proposed and applied to the selection of landslide treatment alternatives. First of all, considering that experts and the public participate in the evaluation of LSGDM events, respectively, the method of obtaining and handling the public evaluation information is given, and the information fusion approach of the public and experts evaluation information is given. Second, evaluation values represented by fuzzy numbers are converted into intuitionistic fuzzy numbers (IFNs), and the intuitionistic fuzzy evaluation decision matrix described by IFNs is obtained. Then, a new LSGDM method of alternatives classification and ranking based on IFS and TWD is proposed, the calculation steps and algorithm description are given. In this process, we first cluster the experts, then consider the identification and management of non-cooperative behavior of expert groups. This work provides an effective method based on LSGDM for the selection of landslide treatment alternatives. Finally, the sensitivity of parameters is analyzed, and the feasibility and effectiveness of this method are compared and verified.

To address the challenges of traffic congestion and suboptimal operational efficiency in the context of large-scale applications like production plants and warehouses that utilize multiple automatic guided vehicles (multi-AGVs), this article proposed using an Improved Q-learning (IQL) algorithm and Macroscopic Fundamental Diagram (MFD) for the purposes of load balancing and congestion discrimination on road networks. Traditional Q-learning converges slowly, which is why we have proposed the use of an updated Q value of the previous iteration step as the maximum Q value of the next state to reduce the number of Q value comparisons and improve the algorithm’s convergence speed. When calculating the cost of AGV operation, the traditional Q-learning algorithm only considers the evaluation function of a single distance and introduces an improved reward and punishment mechanism to combine the operating distance of AGV and the road network load, which finally equalizes the road network load. MFD is the basic property of road networks and is based on MFD, which is combined with the Markov Chain (MC) model. Road network traffic congestion state discrimination method was proposed to classify the congestion state according to the detected number of vehicles on the road network. The MC model accurately discriminated the range near the critical point. Finally, the scale of the road network and the load factor were changed for several simulations. The findings indicated that the improved algorithm showed a notable ability to achieve equilibrium in the load distribution of the road network. This led to a substantial enhancement in AGV operational efficiency.

This study proposes a method of constructing and transforming three-dimensional (3D) models that can convert a 3D model into a chain-type modular configuration and realize the mutual transformation between different configurations with a straight chain as the intermediate state through standard folding steps. A method for detailed representation of voxels is proposed. Based on detailed voxels, an accelerated generation algorithm for the connection forest, which can describe the possible chain configurations, is developed. The foldability verification of the configurations and the generation of the folding operations are realized according to the folding rules. A collision detection algorithm based on encoding and projection is also introduced to detect collisions in the process of folding sequence generation. In this work, an interactive platform is established for users to calculate the input model transformation through simple operations and obtain a simulation animation of the folding operations. The experimental cases prove the effectiveness of the method in constructing and transforming the chain-type modular configurations of the input 3D models.