Data valuation provides a principled framework for quantifying the contribution of data to model training. It plays a crucial role in trustworthy machine learning (ML) by supporting data curation, enhancing interpretability, and enabling fair incentive mechanisms in data markets. Shapley value is a popular method for data valuation, but accurate estimation remains computationally expensive, particularly at the dataset level. In this paper, we introduce Ensemble Shapley, an efficient framework tailored for dataset-level valuation on the sharded structure. To reduce the computational costs, we propose a two-phase estimation method that apportions the intensive contribution computation costs across disjoint data shards and strategically reuses the computation results, achieving efficient contribution evaluation through the ensemble of shard models. However, weak shard models trained on noisy data may degrade ensemble models’ performance. To solve this, we introduce a behavior-driven guided sampling method that pairs noisy datasets with benign ones, ensuring reliable contribution estimates despite the noise. We also derive an advantageous lower bound for the number of evaluation iterations that balances efficiency and accuracy by the number of shards. Experimental results show Ensemble Shapley has superior efficiency over existing methods while maintaining comparable accuracy across various ML tasks, and demonstrates strong scalability and integration potential.

Human-driven vehicles and different levels of autonomous vehicles are expected to coexist on roads in the future. However, autonomous systems behave differently from their human-driver counterparts, these two behaviors are incompatible with one another, thereby negatively impacting traffic efficiency and safety. Herein, we present the construction of human-like trajectories for use in autonomous vehicles as a possible solution to this issue. We present a trajectory planning method based on the spatiotemporal geometric transformation of driving scenarios to generate human-like trajectories. Speed and safety redundancy data were collected through driving tests to understand human driving behaviors. Self-driving scenarios were abstracted as a Lorentz coordinate system under a three-dimensional Minkowski space-time. A surrounding-manifold tensor equation was established using differential geometry theory to depict the relationship between the trajectory constraints and the geometric spatiotemporal background. A metric tensor field can be solved from the equation to construct the corresponding “volcano space-time,” which is a three-dimensional general Riemannian space for placing the subject vehicle and the surroundings. The geodesics of the volcano space-time are solved using the geodesic equation and are projected back into the three-dimensional Minkowski space-time. Geodesic trajectories were fitted as Bézier curves in this study and corrected according to the vehicle dynamics constraints for trackability. In simulation and real vehicle tests, trajectories generated using the proposed algorithm exhibited collision avoidance and trackability, and the algorithm offered behaviors that were similar to those of human drivers under the same scenarios.

Mobile Edge Computing (MEC) is regarded as an enabling technology to provide computing capacity closer to the mobile equipments (MEs). Due to the MEs' mobility, the execution results of offloaded tasks can't be transmitted back to MEs if they are out of the coverage of the small cell network (SCN). Therefore, the migration is needed, which causes the extra cost. In this paper, a novel mobility-aware offloading and migration scheme with MEC is proposed to maximize the total revenue of MEs. Firstly, with the analysis of MEs' mobility, the problem of maximizing the total revenue of MEs is formulated, which integrates the offloading decision and resource allocation to reduce the probability of migration. Secondly, the problem is formulated as a mixed integer non-linear programming (MINLP) problem, so a SUb-optimal algorithm named genetic algorithm based allocation algorithm (GAAA) is proposed to solve this NP-hard problem. Finally, extensive simulations demonstrate that GAAA can effectively improve the total revenue of MEs in contrast with other baseline algorithms.