Open-set semi-supervised learning (OSSL) has attracted growing interest, which investigates a more practical scenario where out-of-distribution (OOD) samples are only contained in unlabeled data. Existing OSSL methods like OpenMatch learn an OOD detector to identify outliers, which often update all modal parameters (i.e., full fine-tuning) to propagate class information from labeled data to unlabeled ones. Currently, prompt learning has been developed to bridge gaps between pre-training and fine-tuning, which shows higher computational efficiency in several downstream tasks. In this paper, we propose a prompt-driven efficient OSSL framework, called OpenPrompt, which can propagate class information from labeled to unlabeled data with only a small number of trainable parameters. We propose a prompt-driven joint space learning mechanism to detect OOD data by maximizing the distribution gap between ID and OOD samples in unlabeled data, thereby our method enables the outliers to be detected in a new way. The experimental results on three public datasets show that OpenPrompt outperforms state-of-the-art methods with less than 1% of trainable parameters. More importantly, OpenPrompt achieves a 4% improvement in terms of AUROC on outlier detection over a fully supervised model on CIFAR10.
翻译:开放的半监督学习(OSSL)吸引了越来越多的人的兴趣,它调查了一个更加实际的情景,即分配外(OOOD)样本只包含在未贴标签的数据中。OpenMatch等现有的OSL方法学习OOD检测器,以识别离线器,这经常更新所有模式参数(即全面微调),以传播从标签数据到未贴标签数据的分类信息。目前,已经发展了迅速学习,以弥合培训前和微调之间的差距,这显示一些下游任务的计算效率更高。在本文中,我们提出了一个快速驱动的高效OSL框架,称为OpenPrompt,它能够传播从标签到未贴标签的数据的类信息,只有少量的可培训参数。我们建议了一个快速驱动的联合空间学习机制,通过在未贴标签数据中的ID和OOOD样本之间的分配差距最大化,从而我们的方法能够以新的方式检测出离线者。三个公共数据集的实验结果显示,Opropropt Prompperoverola proforation arestrain train train 4 a real-trafor-rass arass arass arass arass berass berass berass be trade 4 arefrofrmatial a be truptial aremas be truptial a be be be be be be be be be be be be be truptial 4 4 a be be truptial 4 a rutial- 4 a rutial-rotial-rotial-rotial-rotial-rotimats,我们全面方法在不完全一个低的测试要求充分的路径,在不完全的测试。 4 4 4 aral-rofr-rofrmatial 方法,在不全面的测试性地在4 方法中,在不完全一个不完全的测试性地的测试性地的测试。在4 4 a 4 4 4 4 4 方法中,在4 ad-