ScholarGate
助手
Machine learningDeep learning / NLP / CV

半监督生成对抗网络 (Semi-supervised GAN, SGAN) 将标准的生成对抗网络 (GAN) 判别器进行了扩展,使其能够同时将标记样本分类到 K 个真实类别,并将生成的假样本检测为第 (K+1) 类,从而利用生成器的合成数据作为隐式正则化,并能用极少的标记样本训练出强大的分类器。

在普通的 GAN 中,判别器仅回答“真实还是虚假?”的问题。SGAN 对该判别器进行了重新利用:它现在输出 K+1 个类别,其中前 K 个类别对应于感兴趣的真实类别(如狗、猫等),最后一个类别代表生成的样本。通过同时执行这两项任务,标记样本指导分类器,而数量远多于标记样本的未标记真实数据和生成器的合成样本则可以防止模型对少量标记样本过拟合。其结果是一个分类器,其性能可与在更多标记样本上训练的完全监督模型相媲美。

在 MethodMind 中打开即将推出视频即将推出Download slides

阅读完整方法

仅限会员

使用免费账户登录即可阅读本节。

登录

Method map

The neighbourhood of related methods — select a node to explore.

来源

  1. Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., & Chen, X. (2016). Improved Techniques for Training GANs. Advances in Neural Information Processing Systems (NeurIPS), 29. link
  2. Odena, A. (2016). Semi-Supervised Learning with Generative Adversarial Networks. ICML Workshop on Generative Adversarial Networks. link

如何引用本页

ScholarGate. (2026, June 3). Semi-supervised Generative Adversarial Network. ScholarGate. https://scholargate.app/zh/deep-learning/semi-supervised-gan

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

Compare side by side

被引用于

ScholarGateSemi-supervised GAN (Semi-supervised Generative Adversarial Network). 于 2026-06-15 检索自 https://scholargate.app/zh/deep-learning/semi-supervised-gan · 数据集: https://doi.org/10.5281/zenodo.20539026