DRAUC: An Instance-wise Distributionally Robust AUC Optimization Framework
Authors: Siran Dai, Qianqian Xu, Zhiyong Yang, Xiaochun Cao, Qingming Huang
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, experiments on corrupted benchmark datasets demonstrate the effectiveness of our proposed method. |
| Researcher Affiliation | Academia | Siran Dai1,2 Qianqian Xu3 Zhiyong Yang4 Xiaochun Cao5 Qingming Huang4,3,6 1 SKLOIS, Institute of Information Engineering, CAS 2 School of Cyber Security, University of Chinese Academy of Sciences 3 Key Lab. of Intelligent Information Processing, Institute of Computing Tech., CAS 4 School of Computer Science and Tech., University of Chinese Academy of Sciences 5 School of Cyber Science and Tech., Shenzhen Campus of Sun Yat-sen University 6 BDKM, University of Chinese Academy of Sciences |
| Pseudocode | Yes | See Algorithms 1,2 for more details. |
| Open Source Code | Yes | Code is available at: https://github.com/EldercatSAM/DRAUC. |
| Open Datasets | Yes | For instance, we train our model on binary long-tailed MNIST [22], CIFAR10, CIFAR100 [18], and Tiny-Image Net [21], and evaluate our proposed method on the corrupted version of corresponding datasets [30, 13, 14]. |
| Dataset Splits | Yes | First, we conduct a binary, long-tailed training set. Then, we proceed to train the model on the long-tailed training set with varying imbalance ratios, tune hyperparameters on the validation set, and evaluate the model exhibiting the highest validation AUC on the corrupted testing set. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions PyTorch in its references but does not specify the version of PyTorch or any other software dependencies with version numbers used for the experiments in the main text. |
| Experiment Setup | Yes | First, we conduct a binary, long-tailed training set. Then, we proceed to train the model on the long-tailed training set with varying imbalance ratios, tune hyperparameters on the validation set, and evaluate the model exhibiting the highest validation AUC on the corrupted testing set. [...] The Effect of ϵ. In Figure 2-(a)-(d), we present the sensitivity of ϵ. [...] The Effect of ηλ. In Figure 2-(e)-(h), we present the sensitivity of ηλ. |