Lixin Han

Mainstream semi-supervised learning (SSL) techniques, such as pseudo-labeling and contrastive learning, exhibit strong generalization abilities but lack theoretical understanding. Furthermore, pseudo-labeling lacks the label enhancement from high-quality neighbors, while contrastive learning ignores the supervisory guidance provided by genuine labels. To this end, we first introduce a generalized bias-variance decomposition framework to investigate them. Then, this research inspires us to propose two new techniques to refine them: neighbor-enhanced pseudo-labeling, which enhances confidence-based pseudo-labels by incorporating aggregated predictions from high-quality neighbors; label-enhanced contrastive learning, which enhances feature representation by combining enhanced pseudo-labels and ground-truth labels to construct a reliable and complete symmetric adjacency graph. Finally, we combine these two new techniques to develop an excellent SSL method called GBVSSL. GBVSSL significantly surpasses previous state-of-the-art SSL approaches in standard benchmarks, such as CIFAR-10/100, SVHN, and STL-10. On CIFAR-100 with 400, 2500, and 10,000 labeled samples, GBVSSL outperforms FlexMatch by 3.46%, 2.72%, and 2.89%, respectively. On the real-world dataset Semi-iNat 2021, GBVSSL improves the Top-1 accuracy over CCSSL by 4.38%. Moreover, GBVSSL exhibits faster convergence and enhances unbalanced SSL. Extensive ablation and qualitative studies demonstrate the effectiveness and impact of each component of GBVSSL.

For complex multi-input/output systems, this paper proposes a parallel GRU-LSTM neural network prediction strategy with self-attention mechanism. In order to eliminate the influence between variables, the multidimensional input time series in the system is normalized. The convolution layer is used to extract the initial feature from the sampled data. One branch uses LSTM network to extract and memorize long-term features. The other branch uses GRU network to improve the ability of short-term sequence feature memory. The self-attention mechanism is integrated to realize the strengthening of memory information. The two branches are fused with different weights, and then the predicted value is output through the fully connected layer. The results of comparison with existing network show that the proposed parallel neural network model has better prediction accuracy under the same number of iterations.

The diffusion-based algorithm is a promising member of the family of recommendation algorithms. It makes recommendations through the diffusion process on user-object bipartite graphs. However, a user's taste is often influenced by his/her trusted friends in social networks. In this paper, we propose a new trust-based diffusion on tripartite graphs, which integrates explicit trust relations and implicit trust relations into the diffusion process. Explicit trust relations are obtained from the social networks while implicit trust relations are inferred from implicit feedback. The experimental results indicate that our proposed method has a remarkable improvement in accuracy, and even only implicit trust relations employed, that is, diffusion on user-object bipartite graphs, the recommendation accuracy is still enhanced. We further present a general framework of applying implicit trust relations and explicit trust relations into basic network-based diffusion method, which is more general and flexible, and specific parameters can be selected to meet actual requirements for different datasets and real-world online platforms.

Recommender systems are effective in predicting the most suitable products for users, such as movies and books. To facilitate personalized recommendations, the quality of item ratings should be guaranteed. However, a few ratings might not be accurate enough due to the uncertainty of user behavior and are referred to as natural noise. In this article, we present a novel fuzzy clustering‐based method for detecting noisy ratings. The entropy of a subset of the original ratings dataset is used to indicate the data‐driven uncertainty, and evaluation metrics are adopted to represent the prediction‐driven uncertainty. After the repetition of resampling and the execution of a recommendation algorithm, the entropy and evaluation metrics vectors are obtained and are empirically categorized to identify the proportion of the potential noise. Then, the fuzzy C‐means‐based denoising (FCMD) algorithm is performed to verify the natural noise under the assumption that natural noise is primarily the result of the exceptional behavior of users. Finally, a case study is performed using two real‐world datasets. The experimental results show that our proposal outperforms previous proposals and has an advantage in dealing with natural noise.