Better Embedding and More Shots for Few-shot Learning
Authors: Ziqiu Chi, Zhe Wang, Mengping Yang, Wei Guo, Xinlei Xu
IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | A substantial experimental analysis is carried out to demonstrate the stateof-the-art performance. Compared to the baseline, our method improves by up to 10%+. We also prove that BEMS is suitable for both standard pretrained and meta-learning embedded networks. 4 Experiments 4.1 Experimental Settings Datasets. We perform evaluations on three popular datasets. The mini Image Net [Vinyals et al., 2016] and tiered Image Net [Ren et al., 2018] are the subsets of Image Net. CUB-200-2011 [Wah et al., 2011] is a fine-grained bird classification dataset. All images are resized to 84 84. More details are depicted in Table 1. Evaluation Protocol. We report the average accuracy and 95% confidence interval on random sampled 600 tasks. |
| Researcher Affiliation | Academia | 1Key Laboratory of Smart Manufacturing in Energy Chemical Process, East China University of Science and Technology, China 2School of Information Science and Engineering, East China University of Science and Technology, China |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | Yes | Datasets. We perform evaluations on three popular datasets. The mini Image Net [Vinyals et al., 2016] and tiered Image Net [Ren et al., 2018] are the subsets of Image Net. CUB-200-2011 [Wah et al., 2011] is a fine-grained bird classification dataset. |
| Dataset Splits | Yes | Table 1: Details of datasets. Dataset Classes Images Train/Val/Test mini Image Net 100 60000 64/16/20 tiered Image Net 608 779165 351/97/160 CUB 200 11788 100/50/50 |
| Hardware Specification | No | The paper does not provide specific hardware details used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details with version numbers. It mentions "SGD optimizer" and "Adam optimizer" but no library versions. |
| Experiment Setup | Yes | We train our embedding networks using the standard cross-entropy loss on Cb based on three backbones: Conv Net, Res Net-18, and Wide Res Net. We use SGD optimizer and 128 mini-batch sizes. We use Adam optimizer and an initial learning rate of 0.001. We cut the learning rate in half every 10,000 and 25,000 episodes for mini Image Net and tiered Image Net, respectively. We set the label-smoothing parameter as 0.1. We use 2-layer fully connected layers with Re LU function and a 0.5 dropout as our reconstruction module. The first layer reduces the dimension to half the input dimension, and the second layer restores the dimension. We conduct 200 training iterations. For optimizer, we use Adam with 0.001 learning rate and 0.01 weight decay. We set α = t iter ITER, where iter and ITER are the current and total iterations, respectively. We tune t in {0.01, 0.1, 0.25, 1, 10}. For λ1 and λ2, we tune them in {0.01, 0.02, 0.1, 0.5, 1}. |