Robust Semi-Supervised Learning when Not All Classes have Labels

Authors: Lan-Zhe Guo, Yi-Ge Zhang, Zhi-Fan Wu, Jie-Jing Shao, Yu-Feng Li

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive empirical results show our approach achieves significant performance improvement in both seen and unseen classes compared with previous studies.
Researcher Affiliation Collaboration Lan-Zhe Guo , Yi-Ge Zhang , Zhi-Fan Wu, Jie-Jing Shao, Yu-Feng Li National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China {guolz,zhangyg,wuzf,shaojj,liyf}@lamda.nju.edu.cn. ... This research was supported by the National Key R&D Program of China (2022YFC3340901), the National Science Foundation of China (62176118, 61921006), and the Huawei Cooperation Fund.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes The code is available at https://www.lamda.nju.edu.cn/code_NACH.ashx
Open Datasets Yes We evaluate NACH and compared methods on three SSL benchmark datasets CIFAR-10, CIFAR-100 [18] and Image Net [27].
Dataset Splits No The paper describes how classes are divided (e.g., 50% seen and 50% unseen, then 50% of seen classes as labeled data), but does not provide explicit train/validation/test dataset splits (e.g., percentages or sample counts for each split) needed to reproduce the experiment's data partitioning.
Hardware Specification Yes All experiments are performed on a single NVIDIA 3090 GPU.
Software Dependencies No The paper mentions using Res Net-18 and Res Net-50 as backbone models and Sim CLR, but does not provide specific version numbers for software dependencies (e.g., PyTorch version, Python version).
Experiment Setup Yes For CIFAR datasets, we use Res Net-18 as the backbone model. The model is trained by using the standard Stochastic Gradient Descent method with a momentum of 0.9 and a weight decay of 0.0005. We trained the model for 200 epochs with a batch size of 512.