With the rapid growth of Internet, recommendation systems have become essential for managing information overload. As a classic approach, collaborative filtering has achieved success by leveraging user behavior and item features but struggles with data sparsity and noise. To address these challenges, this paper introduces the Multi-scale Collaborative Diffusion Autoencoder (MCDA). Firstly, the Local Collaborative Relation Encoder captures high-order local interaction patterns. Secondly, to mitigate data sparsity, Multi-scale Collaborative Graph Diffusion based on User-User and Item-Item Similarity incorporates user-user and item-item second-order interactions to introduce multi-level global receptive fields. Thirdly, Global Collaborative Relation Adaptive Filtering Decoder adaptively integrates multi-frequency global signals to reconstruct graph structure, addressing noise introduced during the diffusion process. Finally, the Multi-task Optimization combines reconstruction loss with intra-type contrastive learning to ensure precise graph structure reconstruction while maintaining the discriminability of node embeddings. Comprehensive experiments conducted on multiple public recommendation datasets confirm that MCDA significantly outperforms existing methods, demonstrating its superior effectiveness and robustness.

The online rumor refutation platform is an important channel for clarifying rumors, and the public obtains relevant debunking information and knows the truth through online rumor refutation platforms. Therefore, it is particularly important to reasonably select and utilize relevant online rumor refutation platforms, which can help the public participate in rumor governance and deepen network ecological governance. However, there are many uncertain factors in the entire process of debunking rumors, such as the promotional debunking abilities of different levels of online rumor refutation platforms, while also considering various practical constraints. In order to consider the influence of uncertain factors during the refutation process to improve the effectiveness of refutation, and explore the applicability of uncertainty theory in practice, this paper studies a new mode of refutation information promotion led by social managers such as the government and collaborated by various online refutation platforms, and proposes an uncertain refutation information propagation (URIP) model by considering multiple realistic constraints. Specifically, we characterize the uncertain factors as uncertain variables in the refutation process, such as the debunking ability of online refutation platforms, and solve for the optimal solution by maximizing the expected value of refutation propagation effectiveness. Subsequently, based on the relevant knowledge of uncertainty theory, we derive an equivalent deterministic model for the URIP model. Finally, in order to enable decision-makers to better apply the URIP model in practice, a numerical example is applied to compare the URIP model proposed in this paper with the URIP model without realistic constraints, which proves the superiority of our model.

Sparse R-CNN is a recent strong object detection baseline by set prediction on sparse, learnable proposal boxes and proposal features. In this work, we propose to improve Sparse R-CNN with two dynamic designs. First, Sparse R-CNN adopts a one-to-one label assignment scheme, where the Hungarian algorithm is applied to match only one positive sample for each ground truth. Such one-to-one assignment may not be optimal for the matching between the learned proposal boxes and ground truths. To address this problem, we propose dynamic label assignment (DLA) based on the optimal transport algorithm to assign increasing positive samples in the iterative training stages of Sparse R-CNN. We constrain the matching to be gradually looser in the sequential stages as the later stage produces the refined proposals with improved precision. Second, the learned proposal boxes and features remain fixed for different images in the inference process of Sparse R-CNN. Motivated by dynamic convolution, we propose dynamic proposal generation (DPG) to assemble multiple proposal experts dynamically for providing better initial proposal boxes and features for the consecutive training stages. DPG thereby can derive sample-dependent proposal boxes and features for inference. Experiments demonstrate that our method, named Dynamic Sparse R-CNN, can boost the strong Sparse R-CNN baseline with different backbones for object detection. Particularly, Dynamic Sparse R-CNN reaches the state-of-the-art 47.2% AP on the COCO 2017 validation set, surpassing Sparse R-CNN by 2.2% AP with the same ResNet-50 backbone.

Recent advances in deep neural networks (DNNs) have substantially improved the accuracy and speed of a variety of intelligent applications. Nevertheless, one obstacle is that DNN inference imposes heavy computation burden to end devices, but offloading inference tasks to the cloud causes transmission of a large volume of data. Motivated by the fact that the data size of some intermediate DNN layers is significantly smaller than that of raw input data, we design the DNN surgery, which allows partitioned DNN processed at both the edge and cloud while limiting the data transmission. The challenge is twofold: (1) Network dynamics substantially influence the performance of DNN partition, and (2) State-of-the-art DNNs are characterized by a directed acyclic graph (DAG) rather than a chain so that partition is greatly complicated. In order to solve the issues, we design a Dynamic Adaptive DNN Surgery (DADS) scheme, which optimally partitions the DNN under different network condition. Under the lightly loaded condition, DNN Surgery Light (DSL) is developed, which minimizes the overall delay to process one frame. The minimization problem is equivalent to a min-cut problem so that a globally optimal solution is derived. In the heavily loaded condition, DNN Surgery Heavy (DSH) is developed, with the objective to maximize throughput. However, the problem is NP-hard so that DSH resorts an approximation method to achieve an approximation ratio of 3. Real-world prototype based on self-driving car video dataset is implemented, showing that compared with executing entire the DNN on the edge and cloud, DADS can improve latency up to 6.45 and 8.08 times respectively, and improve throughput up to 8.31 and 14.01 times respectively.

Successful electronic commerce allows businesses to create more efficient channels for product sales or to create new business opportunities. Trust plays an important role in successful online businesses. In this paper, we consider the important case of interactions in peer communities, and then propose a trust establish model based on social computing which aims at avoiding interaction with undesirable participants. Our specific approach based on game to trust computing leads to a decentralized society in which agents help each other weed out undesirable players.