Learning from Noisy Labels with No Change to the Training Process

Authors: Mingyuan Zhang, Jane Lee, Shivani Agarwal

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments confirm our theoretical findings. We conducted two sets of experiments to evaluate our noise-corrected plug-in algorithm.
Researcher Affiliation Collaboration 1Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA, USA 2Twitter, San Francisco, CA, USA (Work done while at the University of Pennsylvania).
Pseudocode Yes Algorithm 1 Noise-Corrected Plug-in Algorithm; Algorithm 3 Iterative Noise Estimation Heuristic
Open Source Code No The paper does not explicitly state that the source code for the proposed methodology is openly available or provide a link to it.
Open Datasets Yes Here we describe experiments on two benchmark data sets, MNIST (Lecun et al., 1998) and CIFAR10 (Krizhevsky & Hinton, 2009)
Dataset Splits No The paper specifies training and test set sizes for synthetic data but does not explicitly detail a separate validation split or its size/methodology. For real datasets, it states it mimicked Patrini et al.'s settings but does not provide specific split details in this paper.
Hardware Specification No No specific hardware details (like GPU or CPU models, memory, or cloud instance types) used for running experiments are provided in the paper.
Software Dependencies No The implementation was in Py Torch (Paszke et al., 2019), and used the Adam W optimizer. No version numbers for PyTorch or Adam W optimizer are provided.
Experiment Setup Yes The optimizer was run for 50 epochs over the training sample; the learning rate parameter was initially set to 0.01 and was halved at the end of every 5 epochs.