Generating High Fidelity Synthetic Data via Coreset Selection and Entropic Regularization

Abstract

Generative models have the ability to synthesize data points drawn from the data distribution, however, not all generated samples are high quality. In this paper, we propose using a combination of coresets selection methods and ``entropic regularization'' to select the highest fidelity samples. We leverage an Energy-Based Model which resembles a variational auto-encoder with an inference and generator model for which the latent prior is complexified by an energy-based model. In a semi-supervised learning scenario, we show that augmenting the labeled data-set, by adding our selected subset of samples, leads to better accuracy improvement rather than using all the synthetic samples.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Pooladzandi et al. "Generating High Fidelity Synthetic Data via Coreset Selection and Entropic Regularization." NeurIPS 2022 Workshops: SyntheticData4ML, 2022.

Markdown

[Pooladzandi et al. "Generating High Fidelity Synthetic Data via Coreset Selection and Entropic Regularization." NeurIPS 2022 Workshops: SyntheticData4ML, 2022.](https://mlanthology.org/neuripsw/2022/pooladzandi2022neuripsw-generating/)

BibTeX

@inproceedings{pooladzandi2022neuripsw-generating,
  title     = {{Generating High Fidelity Synthetic Data via Coreset Selection and Entropic Regularization}},
  author    = {Pooladzandi, Omead and Khosravi, Pasha and Nijkamp, Erik and Mirzasoleiman, Baharan},
  booktitle = {NeurIPS 2022 Workshops: SyntheticData4ML},
  year      = {2022},
  url       = {https://mlanthology.org/neuripsw/2022/pooladzandi2022neuripsw-generating/}
}