摘要
目前,深度学习模型在现实中得到了广泛的应用,当这些模型应用于不同的环境时,可以利用环境中样本分布等经验来进一步提高分类的准确率。基于此,本文提出了基于用户指导的深度学习分类系统DLC-UG,该方法有效地利用了环境中的样本分布来提高分类精度。首先,系统在训练集上训练深度学习模型,接着,用户可以根据自己的经验或他人的建议,在相应的环境中输入每个标签的分布,之后,系统可以通过这个训练模型得到每个样本标签的概率,最后,分布信息有选择地与测试样本上标签的概率相结合。论文在三个真实数据集上选择了六种流行的深度学习模型进行评估,实验结果表明,提出的方法能够提高分类任务的准确率,且明显高于现有的分类方法。
Currently, deep learning models are widely used in many applications. When utilizing these models in different environments, the experience like the distribution of objects in an environment can be used to further increase the accuracy of classification. In this paper, we carried out user guidance based deep learning classification task that is named DLC-UG, which efficiently utilizes this kind of distribution in an environment to increase the accuracy of classification. Firstly, we train the deep learning model on the training set. Secondly, the user can input the distribution of each label in the corresponding environment on the experience of himself or advice from others. Thirdly, we can get the probability of labels on each sample by this trained model. Finally, the distribution information selectively cooperates with the probability of labels on testing samples. We select six popular deep learning models on three real datasets for the evaluation. The experimental results show that our method can increase the accuracy of classification tasks, which is obviously higher than state-of-art methods.
出处
《计算机科学与应用》
2022年第3期707-718,共12页
Computer Science and Application
