Mosaicking to Distill: Knowledge Distillation from Out-of-Domain Data

Authors: Gongfan Fang, Yifan Bao, Jie Song, Xinchao Wang, Donglin Xie, Chengchao Shen, Mingli Song

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

Reproducibility Variable Result LLM Response
Research Type Experimental We validate Mosaic KD over classification and semantic segmentation tasks across various benchmarks, and demonstrate that it yields results much superior to the state-of-the-art counterparts on OOD data.
Researcher Affiliation Collaboration 1Zhejiang University, 2National University of Singapore, 3Central South University 4Alibaba-Zhejiang University Joint Institute of Frontier Technologies
Pseudocode Yes Algorithm 1 Mosaic KD for out-of-domain knowledge distillation
Open Source Code Yes Our code is available at https://github.com/zju-vipa/Mosaic KD.
Open Datasets Yes Four datasets are considered in our experiments as in-domain training set, including CIFAR-100 [26], CUB-200 [53], Stanford Dogs [24] and NYUv2 [34]. For OODKD, we substitute original training data with OOD data, including CIFAR-10 [26], Places365 [63], Image Net [9] and SVHN [36].
Dataset Splits No The paper states, "More details about datasets, training protocol, and metrics can be found in supplementary materials." However, explicit training/test/validation dataset splits (e.g., percentages, sample counts, or explicit reference to predefined splits in the main text) are not provided within the paper's main content.
Hardware Specification No The paper does not explicitly describe the specific hardware used for running its experiments, such as GPU models, CPU types, or cloud computing specifications.
Software Dependencies No The paper does not provide specific version numbers for any software dependencies or libraries used in the experiments.
Experiment Setup No The paper mentions "training protocol" details are in supplementary materials but does not provide specific hyperparameters (e.g., learning rate, batch size, number of epochs, optimizer settings) or other explicit configuration steps for the experimental setup in the main text.