Zero-Shot Learning With Attribute Selection
Authors: Yuchen Guo, Guiguang Ding, Jungong Han, Sheng Tang
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on four ZSL benchmarks demonstrate that our approach can improve zeroshot classification accuracy and yield state-of-the-art results. |
| Researcher Affiliation | Academia | School of Software, Tsinghua University, Beijing 100084, China School of Computing and Communications, Lancaster University, Lancaster, LA1 4YW, UK Institute of Computing Technology, Chinese Academy of Sciences, Beijing, 100190, China |
| Pseudocode | No | The paper describes mathematical optimization steps but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any explicit statement about releasing source code or provide a link to a code repository for the described methodology. |
| Open Datasets | Yes | We use Aw A2 (Xian et al. 2017), a Pascal-a Yahoo (Farhadi et al. 2009), SUN (Patterson and Hays 2012), and CUB (Wah et al. 2011) benchmark datasets, whose statistics are summarized in Tabel 2. We use the datasets, including features, labels, attributes, and data splits, given by: http://www.mpi-inf.mpg.de/zsl-benchmark |
| Dataset Splits | Yes | We use train set and val set which contain source classes and samples for training, and use test set which has target classes and samples for evaluating. When retraining models, as suggested by (Xian et al. 2017), we use train set for training and val set for validation to choose hyper-parameters and use both of them with optimal values for the final model. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using "MATLAB function quadprog" but does not specify any software dependencies with version numbers. |
| Experiment Setup | No | The paper mentions using "val set for validation to choose hyper-parameters", but it does not specify concrete hyperparameter values or other explicit training configurations in the main text. |