Diversity-Driven Synthesis: Enhancing Dataset Distillation through Directed Weight Adjustment

Authors: JIAWEI DU, xin zhang, Juncheng Hu, Wenxin Huang, Joey Tianyi Zhou

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments across multiple datasets, including CIFAR, Tiny-Image Net, and Image Net-1K, demonstrate the superior performance of our method, highlighting its effectiveness in producing diverse and representative synthetic datasets with minimal computational expense.
Researcher Affiliation Collaboration 1 Centre for Frontier AI Research (CFAR), Agency for Science, Technology and Research (A*STAR), Singapore 2 Institute of High Performance Computing, Agency for Science, Technology and Research (A*STAR), Singapore 3Xi Dian University, Xi an, China 4National University of Singapore, Singapore 5Hubei University, Wu Han, China
Pseudocode Yes The pseudocode of our proposed DWA can be found in Algorithm 1.
Open Source Code Yes Our code is available at https://github.com/Angus Dujw/Diversity-Driven-Synthesis.
Open Datasets Yes To evaluate the effectiveness of the proposed method, we have conducted extensive comparison experiments with SOTA methods on various datasets including CIFAR-10/100 (32 32, 10/100 classes) [16], Tiny-Image Net (64 64, 200 classes) [18], and Image Net-1K (224 224, 1000 classes) [3]
Dataset Splits No The paper mentions evaluating on the "original validation set" but does not explicitly provide the specific percentages or counts for training, validation, and test splits (e.g., "80/10/10 split") for the datasets used to reproduce the data partitioning.
Hardware Specification Yes We conduct our experiments on the server with one Nvidia Tesla A100 40GB GPU.
Software Dependencies No Optimizer Adam with {β1, β2} = {0.5, 0.9} Optimizer Adam W with weight decay of 0.01. The paper mentions optimizers but does not list specific versions for software libraries like PyTorch, TensorFlow, or Python itself.
Experiment Setup Yes More detailed hyper-parameter settings can be found in Appendix A.2.1.