Learning Deep Neural Networks under Agnostic Corrupted Supervision

Authors: Boyang Liu, Mengying Sun, Ding Wang, Pang-Ning Tan, Jiayu Zhou

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on multiple benchmark datasets have demonstrated the robustness of our algorithm under different types of corruption. We perform experiments on both regression and classification tasks with corrupted supervision on multiple benchmark datasets. We have performed our experiments on various benchmark regression and classification datasets.
Researcher Affiliation Academia Department of Computer Science and Engineering, Michigan State University, USA.
Pseudocode Yes Algorithm 1 (PRL(G)) Provable Robust Learning for General Corrupted Data. Algorithm 2 (PRL(L)) Efficient Provable Robust Learning for Corrupted Supervision.
Open Source Code Yes Our code is available at https: //github.com/illidanlab/PRL.
Open Datasets Yes For regression, we evaluated our method on the Celeb A dataset, which contains 162,770 training images, 19,867 validation images, and 19,962 test images. We perform our experiments on the CIFAR10 and CIFAR100 datasets to illustrate the effectiveness of our algorithm in classification setting.
Dataset Splits Yes For regression, we evaluated our method on the Celeb A dataset, which contains 162,770 training images, 19,867 validation images, and 19,962 test images. We use a three-layer CNN to train 162770 training images to predict clean coordinates (we use 19867 validation images to do the early stopping). We further split the data into 80% training and 20% test sets.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU models, CPU types, memory amounts) used for running the experiments.
Software Dependencies No The paper mentions using a '9-layer convolutional neural network' but does not specify any software names with version numbers (e.g., PyTorch 1.9, TensorFlow 2.x) that would be needed for reproducibility.
Experiment Setup No The paper does not provide specific hyperparameters like learning rates, batch sizes, or optimizer settings. It mentions 'learning rate γt' in the algorithms, but no concrete value is given in the text.