![](https://img-blog.csdnimg.cn/20201014180756926.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
读顶会
wangdapang_2
这个作者很懒,什么都没留下…
展开
-
2019 CVPR 刘泓 Transferable Adversarial Training: A General Approach to Adapting Deep Classifiers
20年清华特奖得主刘泓的一篇cvpr,人家20岁就发CCF A了原来的对抗学习方法通过“拉近”S、T 之间的距离来实现DA,但是作者指出这种做法会distort 原始的representation。所以本文提出额外生成fs∗f_{s*}fs∗ 、ft∗f_{t*}ft∗(可以理解成代替原来用于adversarial training的特征),然后固定原始feature ,仅调整classifier。“by filling the gap between the source and target d原创 2020-11-28 16:04:04 · 673 阅读 · 1 评论 -
2017nips few-Shot Adversarial Domain Adaptation
2.UDA的两个缺陷首先,数据量小二,不能确保不同domain但是同类别的data的语义对齐3.1 Handling Scarce Target Data问题一: target 的data少,怎么办,扩充它by pairing them with each training source sample.问题二:需要用到target的label信息对策: 创造pairs of samplesdomain-class discriminator DCD 判断是四种pair中的哪一个...原创 2020-11-16 08:23:11 · 250 阅读 · 0 评论 -
2019 ICASSP TRANSFERABLE POSITIVE/NEGATIVE SER VIA CLASS-WISE ADVERSARIAL DOMAIN ADAPTATION
解决监督学习DA问题,和unsupervised 相比,只需要很少的labeled target 数据。是从2018年icasspTowards Conditional Adversarial Training for PredictingEmotions from Speech过来的。相比而言,2018 写得更好,文章中蕴含了很多细节处理,毕竟挂着Björn Schuller舒哥的名字。Introduction中提到文献[6] 不能解决intra-class match的问题。 [6]是 2017 年原创 2020-11-14 09:56:44 · 213 阅读 · 0 评论 -
2018 Interspeech On Enhancing Speech Emotion Recognition using Generative Adversarial Networks
两个GAN网络(不介绍了)两种GAN网络生成,一种粗糙的,一种条件gan。粗糙的不能合成1582维度的,不收敛。条件gan借助了label信息,可以生成1582维的。实验 生成的数据怎么帮助最后的SER准确率three sets of evaluations:(i) in-domain eval- uation合成的样本做训练数据 with and without real data(i...原创 2020-02-11 16:14:17 · 266 阅读 · 0 评论 -
2019 Interspeech CycleGAN-based Emotion Style Transfer as Data Augmentation for SER
利用:Cycle consistent adversarial networks (CycleGAN)目的:addressing the data scarcity problem in speech emotion recognition(1)在 CycleGAN的基础上从大型的unlabeled 语音数据库 迁移特征 到合成的特征表示。(2)扩展了 CycleGAN:用分类loss wh...原创 2020-02-04 21:10:38 · 563 阅读 · 0 评论 -
2019 Interspeech speech emotoin recognition paper reading
2019 Interspeech1. Improved End-to-End Speech Emotion Recognition Using Self Attention Mechanism and Multitask Learning实验2. Self-attention for Speech Emotion Recognition1. Improved End-to-End Speech ...原创 2020-01-08 20:10:21 · 734 阅读 · 3 评论 -
2019 AAAI Transferable Attention for Domain Adaptation
TADA2.Related Works2.1 Adversarial Learning2.2 Global and Local Attention2.3 Transferable Attention for Domain AdaptationDefinition 1 (Transferable Attention)Transferable Local Attention没写完2.Related...原创 2020-01-06 10:51:40 · 1122 阅读 · 3 评论 -
2018 NIPS 条件对抗领域自适应 Conditional Adversarial Domain Adaptation
Conditional Adversarial Domain Adaptation(CADA)因为gpu被沾满了,来解读一下CADA吧。看题目,最先想到的是怎么实现“条件”的,如何刻画“条件”,啥意思呢?就是利用label prediction的信息。分类器GGG 、特征提取器FFF、以及领域判别器DDD。在DANN(baidu一下参考文献,做TL的都懂)里面,FFF是领域判别器DD...原创 2019-12-02 15:08:19 · 1750 阅读 · 1 评论 -
Learning What and Where to Transfer 2019ICML句句解读
Learning What and Where to Transfer句句解读先看meta learning摘要介绍第三段第四段贡献2.L2WWT2.1 Weighted Feature MatchingWhatWhere2.2 Training Meta-Networks and Target Model3. Experiments3.1. Setups3.1 Setups32*32224*22...原创 2019-11-26 17:15:19 · 1901 阅读 · 1 评论