Posterior Regularized Bayesian Neural Network Incorporating Soft and Hard Knowledge Constraints

Published in Knowledge-Based Systems, 2022 · arXiv

Representative figure for this publication
Posterior predictive comparison: with the knowledge constraint (blue), predictions in the unobserved region (shaded band) respect the known conditional range; without it (black), the posterior drifts outside the admissible values.

Injects soft and hard domain-knowledge constraints into Bayesian neural networks through posterior regularization.

Representative publication of this research line.

Recommended citation

@article{huang2022posterior,
  title = {Posterior Regularized Bayesian Neural Network Incorporating Soft and Hard Knowledge Constraints},
  author = {Huang, J. and Pang, Y. and Zhao, X. and Liu, Y. and Yan, H.},
  journal = {Knowledge-Based Systems},
  volume = {249},
  pages = {109047},
  year = {2022},
  eprint = {2210.08608},
  archivePrefix = {arXiv},
  primaryClass = {cs.AI},
  url = {https://arxiv.org/abs/2210.08608}
}