Hyperparameter Learning via Distributional Transfer
Abstract
Bayesian optimisation is a popular technique for hyperparameter learning but typically requires initial exploration even in cases where similar prior tasks have been solved. We propose to transfer information across tasks using learnt representations of training datasets used in those tasks. This results in a joint Gaussian process model on hyperparameters and data representations. Representations make use of the framework of distribution embeddings into reproducing kernel Hilbert spaces. The developed method has a faster convergence compared to existing baselines, in some cases requiring only a few evaluations of the target objective.
Cite
Text
Law et al. "Hyperparameter Learning via Distributional Transfer." Neural Information Processing Systems, 2019.Markdown
[Law et al. "Hyperparameter Learning via Distributional Transfer." Neural Information Processing Systems, 2019.](https://mlanthology.org/neurips/2019/law2019neurips-hyperparameter/)BibTeX
@inproceedings{law2019neurips-hyperparameter,
title = {{Hyperparameter Learning via Distributional Transfer}},
author = {Law, Ho Chung and Zhao, Peilin and Chan, Leung Sing and Huang, Junzhou and Sejdinovic, Dino},
booktitle = {Neural Information Processing Systems},
year = {2019},
pages = {6804-6815},
url = {https://mlanthology.org/neurips/2019/law2019neurips-hyperparameter/}
}