当标注样本匮乏时,半监督学习利用大量未标注样本解决标注瓶颈的问题,但由于未标注样本和标注样本来自不同领域,可能造成未标注样本存在质量问题,使得模型的泛化能力变差,导致分类精度下降.为此,基于wordMixup方法,提出针对未标注样本进行数据增强的u-wordMixup方法,结合一致性训练框架和Mean Teacher模型,提出一种基于u-wordMixup的半监督深度学习模型(semi-supervised deep learning model based on u-wordMixup,SD-uwM).该模型利用u-wordMixup方法对未标注样本进行数据增强,在有监督交叉熵和无监督一致性损失的约束下,能够提高未标注样本质量,减少过度拟合.在AGNews、THUCNews和20 Newsgroups数据集上的对比实验结果表明,所提出方法能够提高模型的泛化能力,同时有效提高时间性能.
随着消费者的健康意识增强,果酒类产品越来越受欢迎。其中发酵型果酒是指利用水果或果汁,经过发酵或半发酵制成的低度酒,含有抗氧化等有益成分,有助于预防心脑血管疾病。本文结合目前国内外酿造型果酒的研究现状,综述了发酵菌株、温度、pH值、溶解氧含量、二氧化碳浓度、菌体密度和基质浓度等因素对果酒品质的影响,探讨了果酒发酵过程中醇类、酸类、酯类及氨基酸类香气成分的生成机制。最后基于果酒酿造产业当前面临的问题,对未来的研究方向和发展趋势进行分析与展望,为后续提高酿造型果酒品质提供理论参考。With the growing awareness of health among consumers, fruit wine products are gaining increasing popularity. Fermented fruit wine, in particular, refers to a type of low-alcohol beverage made by fermenting or partially fermenting fruits or fruit juice. These wines contain beneficial components, such as antioxidants, which can help prevent cardiovascular diseases. This paper reviews the current state of research on fermented fruit wine production both domestically and internationally, summarizing the effects of factors such as fermentation strains, temperature, pH, dissolved oxygen levels, carbon dioxide concentration, cell density, and substrate concentration on fruit wine quality. It further explores the formation mechanisms of aromatic compounds, including alcohols, acids, esters, and amino acids, during the fermentation process. Finally, this study analyzes and forecasts future research directions and development trends in light of the challenges currently facing in the fruit wine brewing industry, providing a theoretical reference for improving the quality of fermented fruit wines.
暂无评论