Dengfeng Liu

Deep neural networks have achieved impressive performance in image classification tasks. However, due to limitations in hardware resources, including computing units and storage capacity, deploying these networks directly on resource-constrained devices such as mobile and edge devices is challenging. While lightweight network models have made significant advancements, the downsampling stage has received little attention. As the feature map is reused multiple times, the reduction of its size during the downsampling stage not only reduces the computational cost of the downsampling module itself but also lowers the computational burden of subsequent stages. This paper addresses this gap by proposing a paddingfree downsampling module that effectively reduces computational costs and can seamlessly integrates into various deep learning models. Furthermore, we introduce a hybrid stem layer to obtain competitive accuracy. Extensive experiments were conducted on CIFAR-100, Stanford Dogs, and ImageNet datasets. On the CIFAR-100 dataset, the results show that the proposed module reduces computational costs by approximately 20% and improves inference speed on resource-constrained devices by around 10%.

Session-based recommendation aims to generate personalized recommendations based on user behaviors within a browsing session. The traditional graph neural network (GNN) methods focus on modeling pair-wise relationships but ignore neighborhood and category relationships which is crucial for improving the accuracy of recommendations. This paper proposes a hypergraph framework based on neighborhood and category association relationships for session-based recommendation (HLNCA). In detail, we first construct a hypergraph that can acquire the neighborhood and category association relationships among items; then aggregate hyperedge representation to obtain the neighborhood and category relationships of nodes; and finally obtain the users' preference through the attention mechanism to calculate the predicted score of items in the candidate set according to the preference. The proposed HL-NCA model outperforms baseline models in terms of Recall@20 and MRR@20 evaluation metrics.

In this paper, the design model and dynamic process of a foldable creative flat folding table are studied. By transforming the problem into a plan view, the coordinates of the edge of the table are written by the geometric relationship, and the multi-objective optimization model is established to determine the optimal folding angle of the folding table, designing processing parameters such as steel bar position and slot length, using Matlab software to draw a dynamic change diagram, so that the production of folding tables to meet customer needs. After analysis and testing, it is believed that the above model design established in this paper can be extended to real life, and can be close to the shape desired by the customer to the greatest extent, and has strong practicability.

Due to the complexity and uncertainty of microbial fermentation processes, data coming from the plants often contain some outliers. However, these data may be treated as the normal support vectors, which always deteriorate the performance of soft sensor modeling. Since the outliers also contaminate the correlation structure of the least square support vector machine (LS-SVM), the fuzzy pruning method is provided to deal with the problem. Furthermore, by assigning different fuzzy membership scores to data samples, the sensitivity of the model to the outliers can be reduced greatly. The effectiveness and efficiency of the proposed approach are demonstrated through two numerical examples as well as a simulator case of penicillin fermentation process.

Production of high quality Chinese rice wine largely depends on fermentation temperature. However, there is no report on the ethanol, sugars, and acids kinetics in the fermentation mash of Chinese rice wine treated at various temperatures. The effects of fermentation temperatures on Chinese rice wine quality were investigated. The compositions and concentrations of ethanol, sugars, glycerol, and organic acids in the mash of Chinese rice wine samples were determined by HPLC method. The highest ethanol concentration and the highest glycerol concentration both were attained at the fermentation mash treated at 23°C. The highest peak value of maltose (90 g/L) was obtained at 18°C. Lactic acid and acetic acid both achieved maximum values at 33°C. The experimental results indicated that temperature contributed significantly to the ethanol production, acid flavor contents, and sugar contents in the fermentation broth of the Chinese rice wines.