Negative Sampling in Semi-Supervised learning
Authors: John Chen, Vatsal Shah, Anastasios Kyrillidis
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on the CIFAR10, CIFAR100, SVHN and STL10 benchmark datasets. Finally, we perform an ablation study for NS3L regarding its hyperparameter tuning. |
| Researcher Affiliation | Academia | 1Department of Computer Science, Rice University, Houston, Texas USA 2Department of Electrical and Computer Engineering, University of Texas at Austin, Austin, Texas USA. Correspondence to: John Chen <johnchen@rice.edu>. |
| Pseudocode | Yes | In this section, we provide the pseudo-code for the Negative Sampling with Semi-Supervised Learning (NS3L) algorithm in Algorithm 1. |
| Open Source Code | No | The paper mentions using existing codebases (Berthelot et al., 2019) and (Oliver et al., 2018) for experiments, but does not provide an explicit statement or link for the open-sourcing of their own NS3L implementation. |
| Open Datasets | Yes | We conduct extensive experiments on the CIFAR10, CIFAR100, SVHN and STL10 benchmark datasets. |
| Dataset Splits | Yes | We use the standard training data/validation data split for SVHN, with 65,932 training images and 7,325 validation images. Similarly, we use the standard training/data validation data split for CIFAR10, with 45,000 training images and 5,000 validation images. We also use the standard data split for CIFAR100, with 45,000 training images and 5,000 validation images. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU/CPU models, memory, or processor types used for running its experiments. |
| Software Dependencies | No | The paper refers to using existing codebases from other works, but it does not specify software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The optimizer is the Adam optimizer (Kingma & Ba, 2014). The batch size is 100, half of which are labeled and half are unlabeled. ... we make the only changes of reducing the total iterations to 200,000, warmup period (Tarvainen & Valpola, 2017) to 50,000, and iteration of learning rate decay to 130,000. ... For NS3L, we use a threshold T = 0.04, learning rate of 6e-4, and λ1 = 1. ... For Mix Match, as in (Berthelot et al., 2019), we use α = 0.75 and λ3 = 75. |