GearNet: Stepwise Dual Learning for Weakly Supervised Domain Adaptation
Authors: Renchunzi Xie, Hongxin Wei, Lei Feng, Bo An8717-8725
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive experimental results show that the performance of existing methods can be significantly improved by equipping with our Gear Net. To verify the effectiveness of our proposed Gear Net, we conduct extensive experiments on widely used benchmark datasets, and experimental results demonstrate that our Gear Net can significantly improve the performance of existing robust methods. |
| Researcher Affiliation | Academia | Renchunzi Xie1, Hongxin Wei1*, Lei Feng2 and Bo An1 1 School of Computer Science and Engineering, Nanyang Technological University, Singapore 2College of Computer Science, Chongqing University, China |
| Pseudocode | Yes | Algorithm 1: Gear Net s Learning Strategy |
| Open Source Code | Yes | The code is published on https://github.com/Renchunzi-Xie/ Gear Net.git |
| Open Datasets | Yes | We simulate experiments based on Office-31 (Saenko et al. 2010) and Office-Home (Venkateswara et al. 2017). |
| Dataset Splits | No | The paper discusses training on source data and pseudo-labeled target data, and evaluates on the target domain, but does not explicitly mention a separate validation set or its split details for hyperparameter tuning. |
| Hardware Specification | Yes | conduct all the experiments on NVIDIA Tesla V100 GPU. |
| Software Dependencies | No | implement all methods by Py Torch - PyTorch is mentioned, but no version number is provided for reproducibility. |
| Experiment Setup | Yes | For comparability, all the experiments use Stochastic gradient descent optimizer with an initial learning rate of 0.003 and a momentum of 0.9. The batch size is set as 32 and the total number of epochs is 200. For Gear Net, the total number of steps is set as 10. |