A Dual Attention Network with Semantic Embedding for Few-Shot Learning
Authors: Shipeng Yan, Songyang Zhang, Xuming He9079-9086
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate our model on three few-shot image classification datasets with extensive ablative study, and our approach shows competitive performances over these datasets with fewer parameters. |
| Researcher Affiliation | Academia | Shipeng Yan, Songyang Zhang, Xuming He School of Information Science and Technology, Shanghai Tech University {yanshp, zhangsy1, hexm}@shanghaitech.edu.cn |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | For facilitating the future research, code and data split are available: https://github.com/tonysy/STANet-Py Torch |
| Open Datasets | Yes | We evaluate our STANet method on the task of few-shot image classification by conducting a set of experiments on three datasets. In addition to two publicly-available datasets, Mini Image Net (Krizhevsky, Sutskever, and Hinton 2012) and Omniglot (Lake, Salakhutdinov, and Tenenbaum 2015), we propose a new few-shot learning benchmark using real-world images from CIFAR100 (Krizhevsky and Hinton 2009), which is referred to as Meta-CIFAR100 dataset. |
| Dataset Splits | Yes | We adopted the splits proposed by (Vinyals et al. 2016; Ravi and Larochelle 2017) with 64 classes for training, 16 for validation, 20 for testing in the meta-learning setting. |
| Hardware Specification | No | The paper does not explicitly describe the hardware used for running its experiments (e.g., specific GPU/CPU models, memory details). |
| Software Dependencies | No | The paper mentions PyTorch implicitly through a GitHub link containing "Py Torch", but does not specify its version number or any other software dependencies with version numbers. |
| Experiment Setup | No | Details of network architecture and experiments configuration are listed in the supplementary material, not explicitly in the main paper text. |