arrow
返回

Deep kernel methods learn better: from cards to process optimization

delete2024-01-19
delete5
delete
OA
AI
M
Mani Valleti *
R
Rama K. Vasudevan
M
Maxim Ziatdinov
S
Sergei V. Kalinin *
DOI:10.1088/2632-2153/ad1a4fdelete
delete原文链接
delete分享
delete收藏
摘要

摘要

En 中文
The ability of deep learning methods to perform classification and regression tasks relies heavily on their capacity to uncover manifolds in high-dimensional data spaces and project them into low-dimensional representation spaces. In this study, we investigate the structure and character of the manifolds generated by classical variational autoencoder (VAE) approaches and deep kernel learning (DKL). In the former case, the structure of the latent space is determined by the properties of the input data alone, while in the latter, the latent manifold forms as a result of an active learning process that balances the data distribution and target functionalities. We show that DKL with active learning can produce a more compact and smooth latent space which is more conducive to optimization compared to previously reported methods, such as the VAE. We demonstrate this behavior using a simple cards dataset and extend it to the optimization of domain-generated trajectories in physical systems. Our findings suggest that latent manifolds constructed through active learning have a more beneficial structure for optimization problems, especially in feature-rich target-poor scenarios that are common in domain sciences, such as materials synthesis, energy storage, and molecular discovery. The Jupyter Notebooks that encapsulate the complete analysis accompany the article.
Keyword:
deep kernel learning
process optimization
latent space
high-dimensional optimization
Bayesian optimization
ferroelectric kinetic model

期刊

M
Machine Learning-Science and Technology
IF:
4.6
论文数:
1.1K
被引数:
4.1K

机构

U
university of tennessee knoxville
学者数:
146
论文数: 82
被引数: 8.4K
University of Tennessee System 封面图
University of Tennessee System
学者数:
2.9W
论文数: 2.7W
被引数: 3.6W
O
oak ridge national laboratory
学者数:
1.6K
论文数: 616
被引数: 1.1W
学者 查看更多机构
引用论文

引用论文

Quantifying the performance of machine learning models in materials discovery
err2023-01-01
err16
errOAAI
errBorg, Christopher K. H.; Muckley, Eric S.; Nyby, Clara; Saal, James E.; Ward, Logan; Mehta, Apurva; Meredig, Bryce
err分享
err收藏
Higher ambulatory blood pressure relates to enlarged Virchow-Robin spaces in first-ever lacunar stroke patients
err2012-07-07
err0
errOAAI
errPim Klarenbeek; Robert J. Oostenbrugge; Jan Lodder; Rob P. W. Rouhl; Iris L. H. Knottnerus; Julie Staals
err分享
err收藏
Sequential Effects of Dioctyltin Dichloride on the Rat Thymus
err1982-01-01
err0
PREAI
errKlara Miller; Mary P. Scott; J. R. Foster
err分享
err收藏
Nutrient stress arrests tentacle growth in the coral model Aiptasia
err2019-02-04
err0
errOAAI
errNils Rädecker; Jit Ern Chen; Claudia Pogoreutz; Marcela Herrera; Manuel Aranda; Christian R. Voolstra
err分享
err收藏
Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries
err2022-04-12
err43
errOAAI
errZiatdinov, Maxim A.; Liu, Yongtao; Morozovska, Anna N.; Eliseev, Eugene A.; Zhang, Xiaohang; Takeuchi, Ichiro; Kalinin, Sergei, V
err分享
err收藏
学者 查看更多内容