迁移学习(EADA)《Unsupervised Energy-based Adversarial Domain Adaptation for Cross-domain Text Classificat

Posted Blair

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了迁移学习(EADA)《Unsupervised Energy-based Adversarial Domain Adaptation for Cross-domain Text Classificat相关的知识,希望对你有一定的参考价值。

论文信息

论文标题:Unsupervised Energy-based Adversarial Domain Adaptation for Cross-domain Text Classification
论文作者:Han Zou, Jianfei Yang, Xiaojian Wu
论文来源:ACL 2021
论文地址:download 
论文代码:download
引用次数:

1 前言

   Energy-based Generative Adversarial Network 结合 DANN 的模型;

2 方法

  整体框架:

  

  $\\textDANN$ 训练目标:

    $\\beginarrayl\\undersetG_f, G_y\\textmin \\quad \\mathcalL_y\\left(\\mathbfX_s, Y_s\\right)-\\gamma \\mathcalL_f\\left(\\mathbfX_s, \\mathbfX_t\\right) \\\\\\undersetG_d\\textmin \\quad \\mathcalL_d\\left(\\mathbfX_s, \\mathbfX_t\\right)\\endarray$

   本文训练目标:

    $\\beginarrayl\\undersetG_f, G_y\\textmin \\quad \\mathcalL_C E\\left(\\mathbfX_s, Y_s\\right)+\\gamma \\mathcalL_A E\\left(\\mathbfX_\\mathbft\\right), \\\\\\undersetG_a\\textmin\\quad \\mathcalL_A E\\left(\\mathbfX_\\mathbfs\\right)+\\max \\left(0, m-\\mathcalL_A E\\left(\\mathbfX_\\mathbft\\right)\\right)\\endarray$

    $\\mathcalL_A E\\left(\\mathbfx_i\\right)=\\left\\|G_a\\left(G_f\\left(\\mathbfx ; \\theta_f\\right) ; \\theta_a\\right)-\\mathbfx_i\\right\\|_2^2$

 

以上是关于迁移学习(EADA)《Unsupervised Energy-based Adversarial Domain Adaptation for Cross-domain Text Classificat的主要内容,如果未能解决你的问题,请参考以下文章

迁移学习(PAT)《Pairwise Adversarial Training for Unsupervised Class-imbalanced Domain Adaptation》

论文笔记 Spatial contrasting for deep unsupervised learning

深度学习 Unsupervised Learning 理论部分

Machine Learning——Unsupervised Learning(机器学习之非监督学习)

如何区分监督学习(supervised learning)和非监督学习(unsupervised learning)

131.007 Unsupervised Learning - Feature Selection | 非监督学习 - 特征选择