Boosting Model Resilience via Implicit Adversarial Data Augmentation
Authors: Xiaoling Zhou, Wei Ye, Zhemg Lee, Rui Xie, Shikun Zhang
IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments across four common biased learning scenarios: long-tail learning, generalized long-tail learning, noisy label learning, and subpopulation shift learning. The empirical results demonstrate that our method consistently achieves state-of-the-art performance, highlighting its broad adaptability. |
| Researcher Affiliation | Academia | 1National Engineering Research Center for Software Engineering, Peking University, China 2Tianjin University, Tianjin, China |
| Pseudocode | Yes | Algorithm 1: Algorithm of Meta-IADA |
| Open Source Code | Yes | The code for Meta-IADA is available in the supplementary materials. |
| Open Datasets | Yes | We experiment across four typical biased learning scenarios, including LT learning, GLT learning, noisy label learning, and subpopulation shift learning, involving image and text datasets. The excluded settings and results (including those on standard datasets) are detailed in the Appendix. The code for Meta-IADA is available in the supplementary materials. 5.1 Long-Tail Learning Four LT image classification benchmarks, CIFAR-LT [Cui et al., 2019], Image Net-LT [Liu et al., 2019], Places-LT [Liu et al., 2019], and i Naturalist (i Nat) 2018 [Jamal et al., 2020], are evaluated. Additionally, two imbalanced text classification datasets are included. Due to space limitations, we only present experiments for CIFAR-LT and i Nat in the main text. Experiments on CIFAR-LT Datasets. |
| Dataset Splits | Yes | To construct metadata, we randomly select ten images per class from the validation data. For the hyperparameters in the IADA loss, α is selected from {0.1, 0.25, 0.5, 0.75, 1}, while keeping β fixed at 1. Additionally, 1,000 images with clean labels are selected from the validation set to compile the metadata. |
| Hardware Specification | No | The training employs the SGD optimizer with a momentum of 0.9 and a weight decay of 5 × 10−4 on a single GPU, spanning 200 epochs. No specific GPU model or other hardware details were provided. |
| Software Dependencies | No | The paper mentions optimizers (SGD, Adam) and models (ResNet, WRN, DistilBert) but does not provide specific version numbers for software libraries or frameworks (e.g., PyTorch, TensorFlow, Python version). |
| Experiment Setup | Yes | We employ the Res Net-32 model [He et al., 2016] with an initial learning rate of 0.1. The training employs the SGD optimizer with a momentum of 0.9 and a weight decay of 5 × 10−4 on a single GPU, spanning 200 epochs. The learning rate is decayed by 0.01 at the 160th and 180th epochs. Additionally, the perturbation network is optimized using Adam, with an initial learning rate of 1 × 10−3. To construct metadata, we randomly select ten images per class from the validation data. For the hyperparameters in the IADA loss, α is selected from {0.1, 0.25, 0.5, 0.75, 1}, while keeping β fixed at 1. |