Bayesian Functional Optimisation with Shape Prior

Pratibha Vellanki, Santu Rana, Sunil Gupta, David Rubin de Celis Leal, Alessandra Sutti, Murray Height, Svetha Venkatesh

AAAI 2019 pp. 1617-1624

doi:10.1609/AAAI.V33I01.33011617 /aaai/2019/vellanki2019aaai-bayesian/

Abstract

Real world experiments are expensive, and thus it is important to reach a target in a minimum number of experiments. Experimental processes often involve control variables that change over time. Such problems can be formulated as functional optimisation problem. We develop a novel Bayesian optimisation framework for such functional optimisation of expensive black-box processes. We represent the control function using Bernstein polynomial basis and optimise in the coefficient space. We derive the theory and practice required to dynamically adjust the order of the polynomial degree, and show how prior information about shape can be integrated. We demonstrate the effectiveness of our approach for short polymer fibre design and optimising learning rate schedules for deep networks.

PDF AAAI Semantic Scholar

Cite

Text

Vellanki et al. "Bayesian Functional Optimisation with Shape Prior." AAAI Conference on Artificial Intelligence, 2019. doi:10.1609/AAAI.V33I01.33011617

Markdown

[Vellanki et al. "Bayesian Functional Optimisation with Shape Prior." AAAI Conference on Artificial Intelligence, 2019.](https://mlanthology.org/aaai/2019/vellanki2019aaai-bayesian/) doi:10.1609/AAAI.V33I01.33011617

BibTeX

@inproceedings{vellanki2019aaai-bayesian,
  title     = {{Bayesian Functional Optimisation with Shape Prior}},
  author    = {Vellanki, Pratibha and Rana, Santu and Gupta, Sunil and de Celis Leal, David Rubin and Sutti, Alessandra and Height, Murray and Venkatesh, Svetha},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2019},
  pages     = {1617-1624},
  doi       = {10.1609/AAAI.V33I01.33011617},
  url       = {https://mlanthology.org/aaai/2019/vellanki2019aaai-bayesian/}
}