In marketing, optimizing subsidy allocation to maximize overall profits is of substantial economic importance. Prior research has employed treatment effect estimation techniques to identify subsidy-sensitive items and design corresponding allocation strategies. However, more accurate treatment effect estimations do not necessarily lead to better allocations, underscoring the critical influence of decision boundaries in decision-making. This paper argues that optimal allocation fundamentally depends on predicting the expected optimal subsidy, a challenge distinct from conventional treatment effect estimation or causal decision-making, which existing approaches fail to address. To fill this gap, we introduce a two-stage Counterfactual optimal subsidy Learning method with an Asymmetric reward (CoLA). In the first stage, we derive a coarse estimate of the expected subsidy threshold by exploiting order information and the conditional independence between expected and observed subsidies. In the second stage, we refine these estimates using an asymmetric loss function, leading to more robust predictions. Under practical budget constraints, we prioritize candidates based on their Sharpe ratios to determine the final subsidy allocation strategy. Experiments on three public datasets and an online A/B test show that our method achieves significant performance improvements, yielding the highest total profit and incremental leverage ratios.

Post-click conversion rate (CVR) prediction is a central task in recommender systems, yet selection bias creates a severe distributional gap between the clicked training samples and the entire inference space. To address selection bias, propensity-based methods such as inverse propensity scoring (IPS) and doubly robust (DR) have been adopted, which aim to estimate the unbiased learning objective from biased training samples. However, these approaches assume strictly positive propensities, implying every user-item pair has a nonzero probability of interaction. In practice, such positivity assumption maybe violated, for example, in food-delivery platforms, some restaurants located more than 10 kilometers away will be blocked for recommendation. In this study, we theoretically show that when such zero-propensity samples, termed extrapolation samples exist, both IPS and DR estimators become biased. To overcome this limitation, we propose ExtraDebias method, which enables debiased recommendation in both non-extrapolation and extrapolation samples. Specifically, we first train a propensity model to identify extrapolation samples with extremely small propensity estimates, then estimate their pseudo-label intervals, and derive an upper bound of the learning objective for extrapolation samples. By minimizing the derived upper bound, debiased learning on extrapolation samples is ensured, while unbiased learning on non-extrapolation samples is achieved by standard IPS. Experiments on four real-world offline datasets and one online A/B test show that ExtraDebias effectively minimizes prediction errors on extrapolation samples and achieves optimal performance.

Currently, the field of image anomaly detection faces significant challenges, primarily due to data scarcity and insufficient model performance. The lack of abnormal samples has made unsupervised methods a research priority; however, the detection performance of unsupervised algorithms typically falls short of that of supervised algorithms. To address this issue, this paper proposes a multi-scale transformer network based on pretrained network features and a reconstruction framework. This approach extracts deep features across multiple levels and performs reconstruction at the feature layer to accurately localize anomalies. Compared to traditional reconstruction-based anomaly detection methods, our approach leverages pretrained feature representations, thus avoiding the generalization issues caused by relying solely on normal samples. Additionally, we introduce a cross-attention module embedded within skip connections, which filters out non-semantic features and enables more refined spatial restoration and feature representation, further reducing the likelihood of overgeneralization and enhancing prediction accuracy. The performance of this method on the industrial dataset MVtecAD and medical datasets demonstrates its strong generalization ability and effective anomaly detection performance.

Magnetic field-driven microrobotic swarms have drawn extensive attention, especially in the field of automatic control. Realizing dynamic path planning and motion control of microrobotic swarms for mobile target tracking is one of the important tasks that still remains unsolved. In this paper, we firstly present an enhanced bidirectional rapidly-exploring random tree star (EB-RRT*) algorithm considering the physical size of the swarm to dynamically plan the optimal path for obstacle avoidance. An image-guided motion controller, which consists of a direction controller and a Genetic Algorithm based Linear Quadratic Regulator (GA-LQR) velocity controller, is then proposed to realize mobile target tracking using microrobotic swarms. Targeted bursting algorithm is subsequently developed to meet the requirement of tracking high-speed (i.e., 20$\mu m/s$) mobile targets. Simulations are performed to validate the proposed methods and obtain the proper ranges of the input parameters for the controllers. Finally, the control effectiveness of mobile target tracking in different conditions and environments is validated by experimental results. Note to Practitioners—The motivation of this work is to develop an effective control scheme for mobile target tracking using microrobotic swarms. Conventional control schemes mainly focus on the control of single microrobots to reach static targets, and thus the desired path is fixed once planned. In addition, the motion of single monolithic microrobots can be modelled precisely. However, in mobile target tracking using microrobotic swarms, dynamic planning algorithms are demanded to frequently update the desired path. Swarms consisting of millions of micro-agents are also difficult to be modelled due to the complex agent-agent interactions. In this work, an effective control scheme consisting of a dynamic path planner, a motion controller and a targeted bursting unit is developed. Real-time dynamic paths will be planned even though the positions of the swarm and the target change rapidly. The precise control of the swarm direction and velocity are achieved, and moreover, using the targeted bursting algorithm, the swarm can be accelerated to approach mobile targets accurately with higher efficiency. Experimental results validates the proposed tracking strategy in different environments with virtual obstacles. The proposed control scheme paves the way for a better understanding of advanced motion control methods for microrobotic swarms.

Control and navigation of microrobotic swarms have drawn extensive attention recently. Avoiding dynamic obstacles using swarms is one of the major challenges that still remain unsolved. In this work, we develop a control strategy to navigate microrobotic swarms to targeted positions while avoiding dynamic obstacles. We first propose a criterion to evaluate the real-time locomotion efficiency during dynamic obstacle avoidance, i.e., the swarm moving direction and the distance between the swarm and the target. Subsequently, a hierarchical radar with three functional boundaries (detection, safety, and prediction circle) is designed for swarms. The optimal moving direction of the swarm with the existence of dynamic obstacles is selected based on the three circles. The effectiveness of the algorithm is validated by simulations and experiments. Using the proposed strategy, the swarm is capable of avoiding multiple moving obstacles and reaching the predefined target. Finally, to show the compatibility of the proposed control method, the swarm is deployed in a micromaze with different dynamic obstacles, and the results also validate the effectiveness of the strategy.

In this paper, we propose a game theoretic approach for joint offloading and resource allocation optimization (JORAO) problem in mobile edge computing (MEC) system. This study not only investigates offloading strategy, but also considers cloud and wireless resource allocation. Specially, the concern of the JORAO problem is to minimize the energy consumption and monetary cost from mobile terminals' perspective. However, the JORAO problem is non-convex and NP hard. Therefore, it is formulated as a JORAO game. The existence of Nash equilibrium (NE) is proved for it. To obtain NE, we also concentrate on cloud and wireless resource allocation algorithm (CWRAA), which is the sub- algorithm of the JORAO game. For the CWRAA, on one hand, we take consideration of OFDM sub-channels allocation and uplink power allocation in radio access networks (RAN). On the other hand, the computation resource allocation in MEC is studied. Simulation results show that the distributed JORAO game algorithm can nearly minimize the total cost of all mobile terminals (MTs) with low complexity. In addition, the energy consumption and completion time are less when the size of data becomes larger compared with existing algorithms.