---
title: [DL輪読会]"CyCADA: Cycle-Consistent Adversarial Domain Adaptation"&"Learning Semantic Representations for Unsupervised Domain Adaptation"
tags:  #deep learning  
author: [Deep Learning JP](https://www.docswell.com/user/DeepLearning2023)
site: [Docswell](https://www.docswell.com/)
thumbnail: https://bcdn.docswell.com/page/YE9N4NWPE3.jpg?width=480
description: 2018/08/03 Deep Learning JP:  http://deeplearning.jp/seminar-2/
published: August 03, 18
canonical: https://www.docswell.com/s/DeepLearning2023/KXYGQX-dlcycada-cycleconsistent-adversarial-domain-adaptationlearning-semantic-representations-for-unsupervised-domain-adaptation
---
# Page. 1

![Page Image](https://bcdn.docswell.com/page/YE9N4NWPE3.jpg)

"CyCADA: Cycle-Consistent Adversarial Domain Adaptation"
DEEP LEARNING JP
& "Learning Semantic Representations for Unsupervised
[DL Papers]
Domain Adaptation" (ICML2018
)
Presentater: Kei Akuzawa, Matsuo Lab. M2
1
http://deeplearning.jp/


# Page. 2

![Page Image](https://bcdn.docswell.com/page/GE8KQKNX7D.jpg)

•
•
•
• ICML2018
• [Hoffman+] CyCADA: Cycle-Consistent Adversarial Domain Adaptation
• [Xie+] Learning Semantic Representations for Unsupervised Domain Adaptation


# Page. 3

![Page Image](https://bcdn.docswell.com/page/LELZXZPNJR.jpg)

: Source
Source
Target
•
Target
:
•
• Source:
• Target:
• Unsupervised Domain Adaptation:
• Supervised Domain Adaptation:
(X_s, Y_s)
(X_t)
(X_t, Y_t)
<-


# Page. 4

![Page Image](https://bcdn.docswell.com/page/4JMVLV4QJW.jpg)

Ganin+ 2016
Source
Target


# Page. 5

![Page Image](https://bcdn.docswell.com/page/PJR2K2WKE9.jpg)

:
•
•
•
•
• Bengio
……(Talk at the ICML'2018 Workshop on Learning with Limited Labels, July 13th, 2018.)
• Current ML theory is strongly dependent on the iid assumption
• Real-life applications often require generalizations in regimes not seen during training
• Humans can project themselves in situations they have never been (e.g. imagine
being on another planet, or going through exceptional events like in many movies)
,


# Page. 6

![Page Image](https://bcdn.docswell.com/page/PEXDLDP57X.jpg)

•
Ganin+ 2016
• Ganin+ 2016
• Tzeng+ 2017
• Saito+ 2018
•
• Taigman+ 2017
• Shrivastava+ 2017
• Hosseini-Asl+ 2018
Shrivastava+ 2017


# Page. 7

![Page Image](https://bcdn.docswell.com/page/3EKXLXDR7D.jpg)

:
• Source
q(y|x)
(
y
!" [$% & ' ] ≠ !* [$* & ' ]
d
)
z
z(
)
p(z|x) p(y|z)
•
z
•
• MMD,
?:
x
z


# Page. 8

![Page Image](https://bcdn.docswell.com/page/L73535RGE5.jpg)

:
:
•
• Discriminator
(z)
• Encoder Discriminator
•
Min-Max
(z)
(z)
Ganin+ 2016


# Page. 9

![Page Image](https://bcdn.docswell.com/page/87DD4DYN7G.jpg)

!!
•
• Fair Prediction:
• Style Transfer:
• Domain Generalization:
(Domain)


# Page. 10

![Page Image](https://bcdn.docswell.com/page/VJPXMX6N78.jpg)

:
• Target
y
•
•
• Unsupervised Image Translation
(X_s)
(X_t^')
(X_t^', y_s)


# Page. 11

![Page Image](https://bcdn.docswell.com/page/2EVL9L3YJQ.jpg)

• Learning Semantic Representations for Unsupervised Domain Adaptation
•
+
(
)
• CyCADA: Cycle-Consistent Adversarial Domain Adaptation
•
+
(
)


# Page. 12

![Page Image](https://bcdn.docswell.com/page/57GXZX3WEL.jpg)

1.
• Title:
Learning Semantic Representations for Unsupervised Domain Adaptation
• Authors:
Shaoan Xie, Zibin Zheng, Liang Chen, Chuan Chen
• Info:
ICML2018 accepted (oral)
• Abstract:
[
[
,
Centroid
]
,
]
Noisy
Noise


# Page. 13

![Page Image](https://bcdn.docswell.com/page/4EQ8L85QJP.jpg)

[
,
]
Saito+ 2018


# Page. 14

![Page Image](https://bcdn.docswell.com/page/KJ42D29YE1.jpg)

:
•
•
H-divergence
(
)


# Page. 15

![Page Image](https://bcdn.docswell.com/page/LE1MZMKN7G.jpg)

:C
(
C)
Semantic Alignment
!!


# Page. 16

![Page Image](https://bcdn.docswell.com/page/GEW393PME2.jpg)

• Semantic Alignment
1.
2.
3.
4.
[saito+ 2017]
(w/
)
(Noisy)
•
•
•
•
Centroid


# Page. 17

![Page Image](https://bcdn.docswell.com/page/47ZV9VKME3.jpg)

Centroid
Centroid
(
)


# Page. 18

![Page Image](https://bcdn.docswell.com/page/YJ65K5Q57V.jpg)

:
• DAN Semantic Loss
• Centroid
•


# Page. 19

![Page Image](https://bcdn.docswell.com/page/GJ5YPYRGJ4.jpg)

• MNIST-USPS-SVHN
• DAN
•


# Page. 20

![Page Image](https://bcdn.docswell.com/page/LE3535PPJ5.jpg)

•
•
•
(Semantic Alignment
Semantic Alignment
[
)
Noisy
Noise
]


# Page. 21

![Page Image](https://bcdn.docswell.com/page/8EDD4DL3EG.jpg)

2.
• Title:
CyCADA: Cycle-Consistent Adversarial Domain Adaptation
• Authors:
Judy Hoffman, Eric Tzeng, Taesung Park, Jun-Yan Zhu, Phillip Isola, Kate Saenko, Alexei
Efros, Trevor Darrell
• Info:
ICML2018 accepted (oral)
• Abstract:
(Pixel)
Domain Translation
CycleGAN


# Page. 22

![Page Image](https://bcdn.docswell.com/page/V7PXMX2PE8.jpg)

•
• Semantic Alignment
•
• c.f.
Bengio, 2017]
• (
)
•
•
• Cycle-GAN
•
• Cycle-GAN
•
(
)
[Jo and


# Page. 23

![Page Image](https://bcdn.docswell.com/page/2JVL9LDV7Q.jpg)

4
•
•
•
•
Component
Cycle-GAN
Semantic Consistency
Classifier
Feature Adaptation


# Page. 24

![Page Image](https://bcdn.docswell.com/page/5EGXZXY1JL.jpg)

: Cycle-GAN
Cycle-GAN
->


# Page. 25

![Page Image](https://bcdn.docswell.com/page/4JQ8L8MN7P.jpg)

: Semantic Consistency Loss
(f_s
pre-train
)


# Page. 26

![Page Image](https://bcdn.docswell.com/page/K742D2V3J1.jpg)

: Classifier


# Page. 27

![Page Image](https://bcdn.docswell.com/page/LJ1MZM3ZEG.jpg)

: Feature Adaptation
Domain Adversarial Network


# Page. 28

![Page Image](https://bcdn.docswell.com/page/GJW393P6J2.jpg)

:
• SVHN -> MNIST
Adaptation
Adaptation
• Pixel
Feature
Pixel
Feature


# Page. 29

![Page Image](https://bcdn.docswell.com/page/4EZV9VKRJ3.jpg)

: Cycle Loss Semantic Loss
• SVHN -> MNIST
Semantic Alignment
• Semantic Segmentation(
Cycle Loss
)
Cycle


# Page. 30

![Page Image](https://bcdn.docswell.com/page/Y765K5Q1EV.jpg)

: Semantic Segmentation


# Page. 31

![Page Image](https://bcdn.docswell.com/page/G75YPYRL74.jpg)

•
• Cycle-GAN
Loss
Cycle Loss
Semantic


# Page. 32

![Page Image](https://bcdn.docswell.com/page/9J2G6GD3JR.jpg)

•
•
•
• [Saito+ 2018b]:
• [Hosseini-Asl+ 2018]:


# Page. 33

![Page Image](https://bcdn.docswell.com/page/DEYV9V28EM.jpg)

References
• Y.Ganin,E.Ustinova,H.Ajakan,P.Germain,H.Larochelle, F. Laviolette, M. Marchand, and V. Lempitsky.
Domain- adversarial training of neural networks. JMLR, 17(59):1–35, 2016.
• Tzeng, E., Hoffman, J., Saenko, K., and Darrell, T. Adversarial discriminative domain adaptation. In
Computer Vision and Pattern Recognition (CVPR), 2017.
• Kuniaki Saito, Yoshitaka Ushiku, Tatsuya Harada. Asymmetric Tri-training for Unsupervised Domain
Adaptation. The 34th International Conference on Machine Learning (ICML 2017), 2017.
• Kuniaki Saito, Kohei Watanabe, Yoshitaka Ushiku, Tatsuya Harada. Maximum Classifier Discrepancy for
Unsupervised Domain Adaptation. The 31th IEEE Computer Society Conference on Computer Vision and
Pattern Recognition (CVPR 2018), 2018, (oral).
• Kuniaki Saito, Shohei Yamamoto, Yoshitaka Ushiku, Tatsuya Harada, Open Set Domain Adaptation by
Backpropagation, arXiv, 2018b
• Taigman, Y., Polyak, A., and Wolf, L. Unsupervised cross-domain image generation. In International
Conference on Learning Representations, 2017a.
• Shrivastava, A., Pfister, T., Tuzel, O., Susskind, J., Wang, W., and Webb, R. Learning from simulated and
unsupervised im- ages through adversarial training. In The IEEE Conference on Computer Vision and Pattern
Recognition (CVPR), July 2017.
• Ehsan Hosseini-Asl, Yingbo Zhou, Caiming Xiong, Richard Socher, Augmented Cyclic Adversarial Learning for
Domain Adaptation, arxiv, 2018
• Jason Jo, Yoshua Bengio, Measuring the tendency of CNNs to Learn Surface Statistical Regularities, arxiv
2017, https://arxiv.org/abs/1711.11561