Learning from Label Proportions by Learning with Label Noise
Authors: Jianxin Zhang, Yutong Wang, Clay Scott
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our approach demonstrates improved empirical performance in deep learning scenarios across multiple datasets and architectures, compared to the leading methods. Experiments with deep neural networks are presented in Section 6, where we observe that our approach outperforms competing methods by a substantial margin. |
| Researcher Affiliation | Academia | Jianxin Zhang, Yutong Wang, and Clayton Scott Electrical Engineering and Computer Science University of Michigan Ann Arbor, MI 48109 {jianxinz, yutongw, clayscot}@umich.edu |
| Pseudocode | Yes | Algorithm 1 LLPFC-ideal, Algorithm 2 LLPFC-uniform, Algorithm 3 LLPFC-approx |
| Open Source Code | Yes | 1Code is available at https://github.com/Z-Jianxin/LLPFC |
| Open Datasets | Yes | We perform experiments on three benchmark image datasets: the letter split of EMNIST [6], SVHN [29], and CIFAR10 [17]. |
| Dataset Splits | No | The paper describes how training bags are generated and references benchmark datasets (e.g., EMNIST, SVHN, CIFAR10). However, it does not provide explicit train/validation/test dataset splits with percentages, sample counts, or citations to predefined splits for these benchmark datasets. |
| Hardware Specification | Yes | All models are trained on a single Nvidia Tesla v100 GPU with 16GB RAM. |
| Software Dependencies | No | The paper mentions 'deep learning scenarios' and 'deep neural networks' and discusses using 'cross-entropy loss', but it does not specify any software frameworks (e.g., PyTorch, TensorFlow) or their version numbers, nor any other libraries with specific versioning information. |
| Experiment Setup | No | The paper states that models were trained on a specific GPU and mentions the setting of weight 'w' and the use of cross-entropy loss. It also says 'Full experimental details are in the appendix'. However, the main text does not provide concrete hyperparameter values like learning rate, batch size, or number of epochs, nor specific optimizer settings beyond referencing 'parameters suggested in the original papers' for network training. |