Multi-Prompt Alignment for Multi-Source Unsupervised Domain Adaptation
Authors: Haoran Chen, Xintong Han, Zuxuan Wu, Yu-Gang Jiang
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that MPA achieves state-of-the-art results on three popular datasets with an impressive average accuracy of 54.1% on Domain Net. |
| Researcher Affiliation | Collaboration | Haoran Chen1,2 Xintong Han3 Zuxuan Wu1,2 Yu-Gang Jiang1,2 1Shanghai Key Lab of Intell. Info. Processing, School of CS, Fudan University 2Shanghai Collaborative Innovation Center of Intelligent Visual Computing 3Huya Inc |
| Pseudocode | No | The paper describes its methodology in prose and through diagrams (Figure 1, Figure 2, Figure 3), but does not provide any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any explicit statements about releasing source code, nor does it provide links to a code repository. |
| Open Datasets | Yes | Experiments are conducted on three popular benchmark datasets of UDA to evaluate the effectiveness of MPA, namely Image CLEF, Office-Home, and Domain Net. [...] Domain Net is the largest dataset to date, consisting of about 0.6 million images from 345 categories in 6 different domains: Clipart, Infograph, Painting, Quickdraw, Real, and Sketch. |
| Dataset Splits | No | The paper describes the datasets used (Image CLEF, Office-Home, Domain Net) and evaluation metrics (top-1 accuracy), but does not explicitly provide details on how these datasets were split into training, validation, and test sets with specific percentages or sample counts for reproducibility, nor does it cite predefined splits directly. |
| Hardware Specification | Yes | The adaptation process with LST requires only 11 hours of GPU time on a NVIDIA RTX 3090, whereas adapting using MPA takes 54 hours a speed increase of approximately 5 times. |
| Software Dependencies | No | The paper mentions using specific models like CLIP and optimizers like mini-batch SGD, but does not provide version numbers for any software dependencies, libraries, or frameworks used in the implementation. |
| Experiment Setup | Yes | Prompts and auto-encoder of MPA are trained using the mini-batch SGD optimizer with a learning rate of 0.003 and 0.005 while the learned subspace is tuned with a 0.0005 learning rate in LST. We use a batch size of 32 and adopt a cosine learning rate scheduler. For hyper-parameters, token lengths M1 and M2 are both set to 16. Pseudo-label threshold τ is set to 0.4 for producing reliable labels. α in Equation 9 is set to 500. The weight matrix W2 of the back projection function in Equation 7 has a size of R384 d I, where d I is 100 for Image CLEF, 150 for Office Home and 250 for Domain Net. |