Looking Wider for Better Adaptive Representation in Few-Shot Learning
Authors: Jiabao Zhao, Yifan Yang, Xin Lin, Jing Yang, Liang He10981-10989
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments for validating our proposed algorithm, which achieves new state-of-the-art performances on two public benchmarks. We have carried out extensive experiments on multiple datasets. Our proposed method outperforms the baseline methods and achieves new state-of-the-art performances. |
| Researcher Affiliation | Collaboration | 1 Shanghai Key Laboratory of Multidimensional Information Processing, ECNU, Shanghai, China 2 School of Computer Science and Technology, East China Normal University, Shanghai, China 3 Transwarp Technology (Shanghai) Co., Ltd, China |
| Pseudocode | No | The paper includes architectural diagrams and mathematical equations but does not contain explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | Mini-Imagenet. ...It contains 100 classes with 600 images per class, which are divided into 64, 16, 20 for training/ validation/ testing. Tiered Imagenet. ...There are 608 classes from 34 super-classes, which are divided into 20, 6, 8 for training/ validation/ testing. |
| Dataset Splits | Yes | Mini-Imagenet. ...It contains 100 classes with 600 images per class, which are divided into 64, 16, 20 for training/ validation/ testing. Tiered Imagenet. ...There are 608 classes from 34 super-classes, which are divided into 20, 6, 8 for training/ validation/ testing. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory specifications) used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details with version numbers (e.g., library names with versions) needed to replicate the experiment. |
| Experiment Setup | Yes | In this work, we adopt the Res Net12 (He et al. 2016) as the backbone for extracting the local features. We train the model in two stages. First, we take an image classification with supervision as the pre-training task for training the Res Net12 (He et al. 2016) with the samples of base classes. This is done by performing an Adaptive Avg Pool2d operation on the features of layer 4 in Res Net12 and input it into a full-connection layer for classifying. Secondly, we train the proposed fewshot method based on the pre-trained backbone following the principle proposed by (Vinyals et al. 2016), where the processes of testing and the training have the same condition. We set λ as 1 in the experiments. |