Domain Adaptation with Invariant RepresentationLearning: What Transformations to Learn?

Repository Structure:

DSAN<br> |└───amazon<br> |    └── dataset (Amazon dataset)<br> |    ├── result<br> |    ├── amazon_utils.py<br> |    ├── dsan.py<br> |    └── flip_gradient.py<br> |    └── logger.py<br> |────imageclef<br> |    └── dataset (ImageCLEF dataset)<br> |    ├── logs<br> |    ├── utils.py<br> |    ├── dsan.py<br> |    └── flip_gradient.py<br>

Instructions on running the code: ##1. Run the following command<br>

# for Amazon
cd amazon
python dsan.py --src $source_domain_name --tgt $target_domain_name 

# for ImageCLEF
cd imageclef
python dsan.py --src $source_domain_name --tgt $target_domain_name

##2. Compute environment for our experiments:<br> CPU: Intel 7700k<br> GPU: GeForce RTX2070<br> 32 GB Memory<br> ##3. Table of the experiment result for Amazon:<br>

Model|B→D|B→E|B→K|D→B|D→E|D→K|E→D|E→B|E→K|K→B|K→D|K→E| ---- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | DIRT-T | 78.6 | 76.1 | 75.5 | 76.8 | 75.2 | 79.1 | 69.6 | 71.0 | 84.2 | 69.2 | 73.3 | 79.5 | MDD | 77.1 | 74.4 | 77.0 | 74.7 | 74.1 | 76.3 | 72.4 | 70.2 | 83.3 |69.3 |73.2 | 82.8 | DSAN | 82.7 | 80.8 | 82.6 | 79.5 | 81.4 | 85.3 | 76.7 | 75.1 | 88.0 | 73.8 | 77.3 | 85.0 |