Rejoining fragment images of precious artifacts is a meaningful task because complete artifacts could provide valuable clues for the research of human civilization. However, existing rejoining methods face several challenges including time-consuming manual annotation, insufficient rejoining accuracy, and prohibitive computation cost. For rejoining fragment images of bone sticks (a precious artifact), we propose a lightweight vision graph neural network called RejoinViG to address these challenges. First, our method avoids time-consuming manual annotation of ballast contour data by experts. Specifically, our method directly takes a pair of fragment images as input and then determines whether the image pair is rejoinable. Second, our method improves rejoining accuracy by contour, script, and texture through dynamically constructing local and global graphs. Third, our method improves rejoining accuracy while reducing computation cost by introducing a new attention mechanism named node self-attention. Extensive experiments demonstrate that our method outperforms the state-of-the-art methods significantly. For example, the Top-1 accuracy of our method is 3.9 times that of SFF-Siam. Surprisingly, our method successfully rejoins a pair of previously unknown but rejoinable fragment images of bone sticks in a real-world scenario.

With the popularity of federated learning, federated domain generalization (FedDG) has attracted more and more attention. Existing works of federated learning indicate that the generalization performance of the global model can be improved when the global model is obtained by aggregating local models according to suitable weights. However, existing methods to calculate weights do not fully utilize the data influences on the global model update, which gives us an opportunity to improve the generalization performance of the global model further. In this paper, we propose the method DI (data influences), which utilizes data influences on the global model update to calculate dynamical weights of local model in each round of training. Specifically, the first component data influence calculator (DIC) of DI calculates local weights of local model from the influences of data on the global model update and we introduce the influence function to complete the calculation process. The second component data influence adjuster (DIA) of DI calculates global weights (which are used in the aggregation process of the global model) from local weights. Extensive experiments indicate that our method improves the generalization performance of models significantly. In particular, our method improves model accuracy on benchmark datasets PACS, OfficeHome, and Office-31 by 1.79%, 1.61%, and 2.39% on average, respectively. Source code is publicly available at github-https://github.com/zikunZHOUHH/Fed-DI.

Differential privacy (DP) is introduced into many fields of AI to preserve privacy. However, introducing DP into AI models is extremely error-prone. To verify whether DP AI models can provide privacy guarantee (quantified by privacy budget) as these models claim, existing methods utilize attack methods to audit whether privacy budget of these models is the same as these models claim. To further improve precision of privacy budget auditing, we propose a brand new way to audit privacy budget, namely directly utilizing the parameters of DP AI models to audit privacy budget. In particular, our method utilizes statistical characteristics variance of the output distribution of DP mechanism to audit privacy budget of DP mechanism. DP AI models are regarded as data samples from output distribution of DP AI model training method and are utilized to approximate the variance of output distribution. The approximated variance is leveraged to estimate the variance of noise distribution of DP mechanism and through the relationship between noise variance and privacy budget, our method calculates the audited privacy budget through estimated noise variance. In addition, to reduce computation overhead, our method constructs parameter selection strategy to identify position whose parameter is suitable for privacy budget auditing. Comprehensive experiments are conducted to verify the effectiveness of our auditing method. Comparison results of five competitive auditing methods demonstrate that our method decreases MAE by 18.29% and decreases MSE by 23.17% on experiment datasets.

Auditing the privacy budget of differential privacy (DP) artificial intelligence (AI) models is necessary to ensure that industrial data are protected at the desired level by DP mechanisms. However, existing auditing methods are not general and precise enough to deal with various kinds of AI models, because the existing auditing methods require customizing audit frameworks and utilize information from model parameters insufficiently. In this article, we propose aprecise andgeneral method toauditthe privacy budget of DPAImodels precisely. Our method associates the parameters of the DP AI model with privacy budget through the Bayesian perspective, achieving tight auditing results with a limited number of DP AI models. Extensive experiments show that our method is more precise and general than existing methods. In particular, the experiments involve ten different datasets, five different models, and three different ways to achieve differential privacy, which indicates the generality of our method. According to empirical experiment results, in 35 out of 36 comparison experiments, our method demonstrates improvements in precision.

Recent work has found that two types of dataset characteristics including the dataset’s coverage and data quality are critical for offline reinforcement learning (RL). To improve the policy, model-based offline RL tries to generate reliable synthetic data to expand the dataset’s coverage based on trained forward and backward dynamics models. However, the characteristic of synthetic data’s quality is ignoring, which raises a question of whether augmenting high-quality synthetic data is efficient for offline RL agents. Motivated by this, we propose a novel forward High-quality Imagination and backward Reliable Check (HIRC), which is an effective data augmentation method to generate high-quality and reliable synthetic data. Specifically, we construct a value-guided forward model to generate high-quality imaginary trajectories, and employ a backward model for reliable checking to obtain synthetic data that better match with pre-collected offline transitions. In other words, the proposed HIRC method can generate high-quality synthetic data on the premise of reliability, which can be combined with model-free offline RL methods. Experimental results on the D4RL benchmark demonstrate that high-quality synthetic data generated by HIRC boosts the performance of a base agent TD3_BC. Especially, HIRC with such a base agent achieves better scores against recent popular model-free and model-based offline RL methods.

With the proposed “carbon peaking” and “carbon neutral” goals in China, the transportation sector, as the second largest consumer of oil and a major producer of greenhouse gases, is a critical area for energy efficiency and emission reduction actions. However, few studies have focused on the effective combination of solving residents' commuting challenges and low-carbon travel. In this paper, by extracting real traffic flow data from taxi and bike-sharing trajectory data, a multilayer complex traffic network is formed to realize an interactive visual exploration of urban traffic patterns. Based on this network a low-carbon travel route recommendation is implemented using a modified genetic algorithm to reduce personal carbon emission and travel costs. Meanwhile, the trip chain level carbon emission estimation method is defined for city streets and recommended routes. With the integration of the above algorithms, a visual analytics system is designed and implemented to support the joint exploration of urban traffic patterns and the street carbon emission distribution, low-carbon mixed traffic route recommendations for inter-community commuting, and optimization of low-carbon recommended routes by adjusting bike stations. Take the taxi and bike-sharing trajectory data in Xiamen, China as an example, an evaluation analysis of the system shows that the method is effective in reducing commuting costs for community residents while reducing personal travel carbon emission.

Bus travel time is uncertain due to the dynamic change in the environment. Passenger analyzing bus travel time uncertainty has significant implications for understanding bus running errors and reducing travel risks. To quantify the uncertainty of the bus travel time prediction model, a visual analysis method about the bus travel time uncertainty is proposed in this paper, which can intuitively obtain uncertain information of bus travel time through visual graphs. Firstly, a Bayesian encoder–decoder deep neural network (BEDDNN) model is proposed to predict the bus travel time. The BEDDNN model outputs results with distributional properties to calculate the prediction model uncertainty degree and provide the estimation of the bus travel time uncertainty. Second, an interactive uncertainty visualization system is developed to analyze the time uncertainty associated with bus stations and lines. The prediction model and the visualization model are organically combined to better demonstrate the prediction results and uncertainties. Finally, the model evaluation results based on actual bus data illustrate the effectiveness of the model. The results of the case study and user evaluation show that the visualization system in this paper has a positive impact on the effectiveness of conveying uncertain information and on user perception and decision making.

Exploring campus lifestyle is conducive to innovating education management, optimizing campus resources allocation, and providing personalized services, but little attention had been paid to the exploration campus lifestyle. A novel interactive system based on behavioral data of campus cards is presented in this paper to provide new ideas and technical support for campus management. Interactive visualization techniques are utilized to help users analyze campus lifestyle via intelligible diagrams. The system contains three functional modules: providing a decision-making reference to educators on students' poverty subsidies, predicting students' academic performance by quantitative analysis, and scheduling cafeteria repast based on the scheduling model during the outbreak of COVID-19. Finally, three exploratory case studies are presented to demonstrate the effectiveness of the system.