WDGRL

KO | EN

gitlite — search

#javascript #react #typescript #llm #android #machine-learning #macos #java #windows #awesome #cli #kubernetes #vue #awesome-list

(★ 130)

No description available.

File Explorer

Download Latest Version (.zip)

- - amazon.iml
  - misc.xml
  - modules.xml
  - workspace.xml
- - books_test.svmlight
  - books_train.svmlight
  - dvd_test.svmlight
  - dvd_train.svmlight
  - electronics_test.svmlight
  - electronics_train.svmlight
  - kitchen_test.svmlight
  - kitchen_train.svmlight
- amazon_baseline.py
- amazon_wd.py
- flip_gradient.py
- tsne.py
- tsne_wd.py
- utils.py
- - coral.py
  - dann.py
  - flip_gradient.py
  - mmd.py
  - source_only.py
  - utils.py
  - wd.py
- - misc.xml
  - modules.xml
  - office-caltech.iml
  - workspace.xml
- data_process.py
- flip_gradient.py
- office_baseline.py
- office_wd.py
- utils.py
- - misc.xml
  - modules.xml
  - toy.iml
  - workspace.xml
- dann.py
- dann_domain_classifier.png
- flip_gradient.py
- non_transfer.py
- target_acc.png
- toy_visualization.png
- utils.py
- utils.pyc
- wd.py
README.md

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# WDGRL

**Tensorflow version: 1.3.0**

This repository contains the experiment code implemented in the paper [Wasserstein Distance Guided Representation Learning for Domain Adaptation](https://arxiv.org/abs/1707.01217). The title of the previous version was "Adversarial Representation Learning for Domain Adaptation" and we realized it was too general and thus not proper for this work.

## Toy experiment

We generate a synthetic dataset to show the superior gradient advantage of WDGRL over DANN. In the paper, we claim that when two representation distibutions are distant or have regions they differ a lot, DANN will have gradient vanishing problem while WDGRL still provide stable gradient. It is a little difficult to control such situations, so we design a rather restrictive experiment. However, this toy experiment does verify DANN may fail in some situations where WDGRL still work. The data input visualization is as follows.

![](https://raw.githubusercontent.com/RockySJ/WDGRL/master/toy/toy_visualization.png)

The code files are in the "toy" fold. To make it simple, we create a python file for each approach: non_transfer.py for S-only, dann.py for DANN, wd.py for WDGRL. When we run the programs, we find that WDGRL can sucessfully train a classifer that works well in the target domain. DANN fails this toy experiment since the domain classifier's loss is minimized to zero and thus there is a gradient vanish problem.

![](https://raw.githubusercontent.com/RockySJ/WDGRL/master/toy/dann_domain_classifier.png)
![](https://raw.githubusercontent.com/RockySJ/WDGRL/master/toy/target_acc.png)

**Discussions about this topic are welcomed.**

## Amazon experiment

In amazon_baseline.py, all compared approaches are implemented. As for the MMD implementation, we mainly follow the [code](https://github.com/tensorflow/models/tree/master/research/domain_adaptation). And the WDGRL code is in the amazon_wd.py.

One can run tensorboard to visualize the training by the following command.
```bash
cd WDGRL/amazon/log
tensorboard --logdir='./'
```

## Office-Caltech experiment

In this repository, we provide three features for Office-Caltech data: Surf, GoogleNet1024, CaffeNet4096. Before we begin the experiment, we should decide which feature to use and how to split the train and test sets. By running the code data_process.py, we can generate the pkl data for each domain. The experiments will be fair if the feature to use and the train/test split are the same for each compared approaches. The results for the new train/test split may be different from the results in the paper. We provide the data we use in our experiments [here](https://drive.google.com/open?id=1apft8Ppw4WmA0SAJgy7cF1PW6-W3njtS) while we also suggest that one can generate his data if he uses a different train/test split or a different evaluation protocol.

## Deep imgae experiments

**MNIST->USPS.**
The code is in the path: digits/mnist_usps/

| Methods         |    Source Only | MMD    | CORAL | DANN | WDGRL |
| :-------- | :--------| :-- | :-- | :-- | :-- |
| Target accuracy |    89.79       |  96.77 | 96.61 |95.91 | 97.04 |

WDGRL

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

WDGRL

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

Download README