$i$-Mix: A Domain-Agnostic Strategy for Contrastive Representation Learning

Authors: Kibok Lee, Yian Zhu, Kihyuk Sohn, Chun-Liang Li, Jinwoo Shin, Honglak Lee

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

Reproducibility Variable Result LLM Response
Research Type Experimental In experiments, we demonstrate that i-Mix consistently improves the quality of learned representations across domains, including image, speech, and tabular data. Furthermore, we confirm its regularization effect via extensive ablation studies across model and dataset sizes.
Researcher Affiliation Collaboration 1University of Michigan 2Amazon Web Services 3Google Cloud AI 4KAIST 5LG AI Research
Pseudocode Yes Algorithm 1 Loss computation for i-Mix on N-pair contrastive learning in Py Torch-like style.
Open Source Code Yes The code is available at https://github.com/kibok90/imix.
Open Datasets Yes CIFAR-10/100 (Krizhevsky & Hinton, 2009) consist of 50k training and 10k test images, and Image Net (Deng et al., 2009) has 1.3M training and 50k validation images...
Dataset Splits Yes CIFAR-10/100 (Krizhevsky & Hinton, 2009) consist of 50k training and 10k test images, and Image Net (Deng et al., 2009) has 1.3M training and 50k validation images...
Hardware Specification No The paper does not provide specific hardware details (e.g., specific GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments, only general model architectures like Res Net-50.
Software Dependencies No The paper mentions 'Py Torch-like style' for Algorithm 1 and adapting 'code for supervised contrastive learning', but does not provide specific software names with version numbers.
Experiment Setup Yes Models are trained with a batch size of 256 (i.e., 512 including augmented data) for up to 4000 epochs on CIFAR-10 and 100, and with a batch size of 512 for 800 epochs on Image Net. For i-Mix, we sample a mixing coefficient λ Beta(α, α) for each data, where α = 1 unless otherwise stated. The temperature is set to τ = 0.2. The memory bank size of Mo Co is 65536 for Image Net and 4096 for other datasets, and the momentum for the exponential moving average (EMA) update is 0.999 for Mo Co and BYOL.