Purify Unlearnable Examples via Rate-Constrained Variational Autoencoders

Authors: Yi Yu, Yufei Wang, Song Xia, Wenhan Yang, Shijian Lu, Yap-Peng Tan, Alex Kot

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate the remarkable performance of our method across CIFAR-10, CIFAR-100, and a 100-class Image Net-subset.
Researcher Affiliation Collaboration 1 Rapid-Rich Object Search Lab, Interdisciplinary Graduate Programme, Nanyang Technological University, Singapore 2 School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 3 Peng Cheng Laboratory, Shenzhen, China 4 School of Computer Science and Engineering, Nanyang Technological University, Singapore.
Pseudocode Yes Algorithm 1 Two-stage purification framework of unlearnable examples with D-VAE
Open Source Code Yes Code is available at https: //github.com/yuyi-sd/D-VAE.
Open Datasets Yes We choose three commonly used datasets: CIFAR-10, CIFAR-100 (Krizhevsky et al., 2009), and a subset of Image Net (Deng et al., 2009) with the first 100 classes.
Dataset Splits No The paper mentions 'clean training dataset T' and 'clean test dataset D', but does not explicitly detail train/validation/test splits with percentages or sample counts for reproducibility. It also does not mention a separate validation set being used for hyperparameter tuning.
Hardware Specification Yes It s important to note that the times are recorded using CIFAR-10 as the dataset, Py Torch as the platform, and a single Nvidia RTX 3090 as the GPU.
Software Dependencies No Py Torch as the platform. (Only 'Py Torch' is mentioned without a version number or other specific software dependencies with versions).
Experiment Setup Yes For CIFAR-10, we use 60 epochs, while for CIFAR-100 and the Image Net, 100 epochs are allowed. In all experiments, we use SGD optimizer with an initial learning rate of 0.1 and the Cosine Annealing LR scheduler, keeping a consistent batch size of 128. For D-VAE training on unlearnable CIFAR-10, we use a KLD target of 1.0 in the first stage and 3.0 in the second stage, with only a single 0.5 downsampling to preserve image quality. For the CIFAR-100, we maintain the same hyperparameters as CIFAR-10, except for setting kld2 to 4.5. For Image Net-subset, which has higher-resolution images, we employ more substantial downsampling ( 0.125) in the first stage and set a KLD target of 1.5, while the second stage remains the same as with CIFAR.