Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Semantic Feature Learning for Universal Unsupervised Cross-Domain Retrieval
Authors: Lixu Wang, Xinyu Du, Qi Zhu
NeurIPS 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments across multiple datasets and scenarios, including close-set, partial, and open-set CDR, demonstrate that our approach significantly outperforms existing state-of-the-art CDR methods and other related methods in solving U2CDR challenges. |
| Researcher Affiliation | Collaboration | Lixu Wang Northwestern University, IL, USA EMAIL Xinyu Du General Motors Global R&D, MI,USA EMAIL Qi Zhu Northwestern University, IL, USA EMAIL |
| Pseudocode | No | The paper describes its proposed framework and methods in narrative text and refers to Figure 1 for an overview. It does not include any explicitly labeled 'Algorithm' or 'Pseudocode' blocks. |
| Open Source Code | No | The paper states 'More implementation details, experiment results, and source codes are provided in the Supplementary Materials.' However, a corresponding note in the NeurIPS checklist (Question 5) states: 'We are checking the relevant regulations of our institutions to release the source code. All the datasets used for this work are publicly available.' This indicates that the code is not yet publicly accessible at the time of publication. |
| Open Datasets | Yes | Datasets. Office-31 [38] includes three domains with 31 classes: Amazon (A), DSLR (D), Webcam (W). Office-Home [39] contains four different domains: Art (A), Clipart (C), Product (P), Real (R). And each domain has 67 data categories. Domain Net [40] is the most challenging cross-domain dataset to our best knowledge, which includes six domains: Quickdraw (Qu), Clipart (Cl), Painting (Pa), Infograph (In), Sketch (Sk) and Real (Re). |
| Dataset Splits | No | The paper specifies training epochs and general experiment settings, but it does not provide explicit details about the split percentages or sizes for training, validation, or test datasets. It mentions training epochs like 'The training epochs of IDSE are 100 for Office-31 and Office-Home, and 200 for Domain Net.' but not data splits. |
| Hardware Specification | No | The paper does not explicitly state specific hardware details such as CPU/GPU models (e.g., 'NVIDIA A100', 'Intel Xeon'), memory, or specific computing environments like cloud instance types used for running its experiments. |
| Software Dependencies | No | The paper mentions general software components like 'Res Net-50', 'Mo Cov2', 'SGD optimizer', and 'K-Means', but it does not provide specific version numbers for these or for core programming languages/libraries (e.g., 'Python 3.8', 'PyTorch 1.9'). |
| Experiment Setup | Yes | The SGD optimizer with a momentum of 0.9 is adopted with an initial learning rate of 0.0002 that is scheduled to zero by a cosine learning strategy. The batch size is 64. The training epochs of IDSE are 100 for Office-31 and Office-Home, and 200 for Domain Net. The epoch number of CDSM is 50 for all three datasets. |