【论文阅读】Universal Domain Adaptation
Universal Domain Adaptation
SUMMARY@2020/3/27
文章目录
- Motivation
- Related Work
- Challenges / Aims /Contribution
- Method Proposed
- Feature extractor FFF
- Label classifier GGG
- **Non-adversarial** domain discriminator D′D^\primeD′
- Adversarial domain discriminator DDD
- Training
- Testing
- Experiment
Motivation
This paper focuses on its special setting of universal domain adaptation, where
- no prior information about the target label sets is provided.
- we know source domain with labeled data.
The following figure shows this motivation of this setting.
and the following show some settings of this universal domain adaptation:
Related Work
This work is partly based on some early works of partially set domain adaptation by Mingsheng Long group, like:
- SAN (Partial Transfer Learning with Selective Adversarial Networks)
- utilizes multiple domain discriminators with class level and instance-level weighting mechanism to achieve per-class adversarial distribution matching.
- PADA (Partial adversarial domain adaptation)
- only one adversarial network and jointly applying class-level weighting on the source classifier
- haven’t yet read
and some of others’ relative work of :
- IWAN (Importance weighted adversarial nets for partial domain adaptation)
- constructs an auxiliary domain discriminator to quantify the probability of a source sample being similar to the target domain.
- haven’t yet read
And these works all partly applies the idea of adversarial network GAN and domain adaptation version GAN:
- GAN (Generative Adversarial Nets)
- DANN( Domain-Adversarial Training of Neural Networks)
- adversarial-based, deep method domain adaptation
Challenges / Aims /Contribution
Under the universal domain adaptation setting, our goal now is to match the common categories in source and target domain. The main challenges of solving this universal problem are:
how to deal with the CSˉ\bar {C_S}CSˉ part which is unrelated part of source domain to circumvent negative transfer for target domain
effective domain adaptation between related part of source domain and target domain
learn model (feature extraction & classifier )to minimize the target risk in the common set CCC
Method Proposed
UAN(universal adaptation network) is composes of 4 parts in training phase as following figure shows.
Feature extractor FFF
- find good features that match source and target
- good features to be used by classifier
Label classifier GGG
compute prediction label y^=G(F(x))∈CS\hat y = G(F(x)) \in C_Sy^=G(F(x))∈CS (source domain label set)
classification loss need to be minimized by good parameters of FFF and GGG
EG=E(x,y)∼pL(y,G(F(x)))E_G = \mathbb E_{(\mathrm{x,y})\sim p}L(\mathrm{y},G(F(\mathrm x))) EG=E(x,y)∼pL(y,G(F(x)))
Non-adversarial domain discriminator D′D^\primeD′
compute similarity of each x\rm xx to source domain
- d^′=D′(z)∈[0,1]\hat d^\prime = D^\prime(\rm z) \in[0,1]d^′=D′(z)∈[0,1]
- $\hat d^\prime \rightarrow 1 $ if x is more similar to source
domain classification loss need to be minimized, thus end up with good d^′\hat d^\primed^′ output for every sample from both source and target domain:
ED′=−Ex∼plog(D′(F(x)))−Ex∼qlog(1−D′(F(x)))E_{D^\prime} = - \mathbb E_{\mathrm{x}\sim p}\mathrm{log}(D^\prime(F(\mathrm x))) - \mathbb E_{\mathrm{x}\sim q}\mathrm{log}(1- D^\prime(F(\mathrm x))) ED′=−Ex∼plog(D′(F(x)))−Ex∼qlog(1−D′(F(x)))hypothesis: expectation of similarity value from different label set distribution will be used in weighting adversarial domain discriminator D:
Ex∼pCSˉd^′>Ex∼pCd^′>Ex∼qCd^′>Ex∼qCtˉd^′\mathbb E_{\mathrm x\sim {p_{\bar {C_S}}}} {\hat d^\prime} > \mathbb E_{\mathrm x\sim {p_{ {C}}}} {\hat d^\prime} > \mathbb E_{\mathrm x\sim {q_{{C}}}} {\hat d^\prime} > \mathbb E_{\mathrm x\sim {q_{\bar {C_t}}}} {\hat d^\prime} Ex∼pCSˉd^′>Ex∼pCd^′>Ex∼qCd^′>Ex∼qCtˉd^′not used in adversarial, since it is the same as in DANN, which aims at matching the exactly same source and target label space. may cause negative transfer in universal setting.
Adversarial domain discriminator DDD
aims at discriminate source and target in the common label set CCC
domain discriminate loss: needs to be minimized for good discriminator; needs to be maximized which equals the good representation of feature extractor:
ED=−Ex∼pws(x)log(D′(F(x)))−Ex∼qwt(x)log(1−D′(F(x)))E_{D} = - \mathbb E_{\mathrm{x}\sim p}w^s(\mathrm x)\mathrm{log}(D^\prime(F(\mathrm x))) - \mathbb E_{\mathrm{x}\sim q}w^t(\mathrm x)\mathrm{log}(1- D^\prime(F(\mathrm x))) ED=−Ex∼pws(x)log(D′(F(x)))−Ex∼qwt(x)log(1−D′(F(x)))add big weights for samples from common label set in both source and target domain , aims at maximally match the source and target domain specially in common label set.
weights(called “sample level transferability criterion”) to be constructed:
Ex∼pCws(x)>Ex∼pˉCsws(x)Ex∼qCwt(x)>Ex∼qˉCtwt(x)\mathbb E_{\mathrm x\sim {p_{{C}}}} w^s(\mathrm x) > \mathbb E_{\mathrm x\sim {\bar p_{{C_s}}}} w^s(\mathrm x) \\ \mathbb E_{\mathrm x\sim {q_{{C}}}} w^t(\mathrm x) > \mathbb E_{\mathrm x\sim {\bar q_{{C_t}}}} w^t(\mathrm x) Ex∼pCws(x)>Ex∼pˉCsws(x)Ex∼qCwt(x)>Ex∼qˉCtwt(x)use entropy of predicted vector to measure uncertainty of prediction:
Ex∼qCtˉH(y^)>Ex∼qCH(y^)>Ex∼pCH(y^)>Ex∼pCsˉH(y^)\mathbb E_{\mathrm x\sim {q_{\bar {C_t}}}} H(\mathrm {\hat y}) >\mathbb E_{\mathrm x\sim {q_{{C}}}} H(\mathrm {\hat y}) >\mathbb E_{\mathrm x\sim {p_{{C}}}} H(\mathrm {\hat y}) >\mathbb E_{\mathrm x\sim {p_{\bar {C_s}}}} H(\mathrm {\hat y}) Ex∼qCtˉH(y^)>Ex∼qCH(y^)>Ex∼pCH(y^)>Ex∼pCsˉH(y^)use domain similarity and the prediction uncertainty of each sample, to develop a weighting mechanism for discovering label sets shared by both domains and promote common-class adaptation
ws(x)=H(y^)log∣Cs∣−d^′(x)wt(x)=d^′(x)−H(y^)log∣Cs∣w^s(\mathrm x) = \frac{H(\mathrm {\hat y})}{\mathrm{log}|C_s|}-\hat d^\prime(\mathrm x) \\ w^t(\mathrm x) = \hat d^\prime(\mathrm x)-\frac{H(\mathrm {\hat y})}{\mathrm{log}|C_s|}\\ ws(x)=log∣Cs∣H(y^)−d^′(x)wt(x)=d^′(x)−log∣Cs∣H(y^)- normalized H
- all together normalized when training
Training
- to write in GAN-based two stage, but in neural network implemented end-to-end by using the gradient reversal layer from DANN:
KaTeX parse error: Expected group after '_' at position 6: \max_̲\limits{D}\min_…
Testing
see figure below :
- no adversarial DDD
- calculate weight wt(x)w^t(x)wt(x)for sample xxx from target
- set a validated threshold to argue whether x comes from common label set
Experiment
- FFF is pretrained ResNet50
- all unknown in target labeled as a whole "unknow " big class
- better than prior setting methods
【论文阅读】Universal Domain Adaptation相关推荐
- [论文阅读] Unsupervised Domain Adaptation for Cross-Modality Retinal Vessel Segmentation
[论文地址] [代码] [ISBI 22] Abstract 各种深度学习模型已经被开发出来,用于从医学图像中分割解剖结构,但当在另一个具有不同数据分布的目标域上测试时,它们通常表现不佳.最近,人们提 ...
- 【论文阅读】Domain Adaptation for Deep Entity Resolution
Domain Adaptation for Deep Entity Resolution 摘要 实体解析(ER)是数据集成的一个核心问题.ER的最先进(SOTA)结果是通过基于深度学习(DL)的方法实 ...
- 阅读笔记:Universal Domain Adaptation through Self-Supervision
paper: https://arxiv.org/pdf/2002.07953.pdf code: https://github.com/VisionLearningGroup/DANCE 参考请引用 ...
- [论文阅读] Unsupervised Domain Adaptive Salient Object Detection Through Uncertainty-Aware Pseudo-Label
论文地址:https://arxiv.org/abs/2202.13170 代码:https://github.com/Kinpzz/UDASOD-UPL 发表于:AAAI 22 Abstract 深 ...
- 【论文阅读】Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift
Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift SUMMARY@ 2020 ...
- 从近年顶会论文看领域自适应(Domain Adaptation)最新研究进展
©PaperWeekly 原创 · 作者 | 张一帆 学校 | 中科院自动化所博士生 研究方向 | 计算机视觉 Domain Adaptation 即在源域上进行训练,在目标域上进行测试. 本文总结了 ...
- 2020 Domain Adaptation 最新论文:插图速览(三)
2020 Domain Adaptation 最新论文:插图速览(三) 目录 Learning to Combine: Knowledge Aggregation for Multi-Source D ...
- 主动领域自适应(Active Domain Adaptation)部分经典论文汇总
只简单找了一小部分 CVPR [link] [code] [TQS] [21] Transferable Query Selection for Active Domain Adaptation [l ...
- 论文笔记:Visual Domain Adaptation with Manifold Embedded Distribution Alignment
论文:Visual Domain Adaptation with Manifold Embedded Distribution Alignment 地址:https://arxiv.org/abs/1 ...
最新文章
- 国家公布人工智能四大平台 AI的春天到了?
- springboot-web开发(请求参数)
- android 无法运行 项目红叉_Android Studio 项目报红叉没法运行的几种解决办法总结...
- Python第四章__装饰器、迭代器
- spingMVC问题小结
- pip install -q git+https://github.com/tensorflow/docs.git报错
- 【慢慢学算法】:奇偶校验(bitset使用)
- Safari、IE8、iPhone和BlackBerry在Pwn2Own竞赛中被挑落
- wpf treeview使用expanded事件出错的问题
- java某些基础知识点整理
- atitit.php 流行框架 前三甲为:Laravel、Phalcon、Symfony2 attilax 总结
- 9008刷机工具_黔隆科技刷机教程OPPOR11忘记密码免刷机保资料解屏幕锁教程
- 京东/拼多多淘客小程序跳入路径
- 对联广告php代码,JS实现很实用的对联广告代码(可自适应高度)
- Unity Gyro之使用陀螺仪实现简单VR全景环视效果
- 在线latex 网址
- 网络请求框架 -- 理解Https
- 【JSD2209-DAY05】for、while、数组(上)
- ALSA子系统(五)------XRUN排查
- 计算机组成原理课内实验,【计算机基础论文】计算机组成原理课程实验教学改革(共2885字)...