Dataset Information

Leveraging Theory for Enhanced Machine Learning.

ABSTRACT: The application of machine learning to the materials domain has traditionally struggled with two major challenges: a lack of large, curated data sets and the need to understand the physics behind the machine-learning prediction. The former problem is particularly acute in the polymers domain. Here we aim to simultaneously tackle these challenges through the incorporation of scientific knowledge, thus, providing improved predictions for smaller data sets, both under interpolation and extrapolation, and a degree of explainability. We focus on imperfect theories, as they are often readily available and easier to interpret. Using a system of a polymer in different solvent qualities, we explore numerous methods for incorporating theory into machine learning using different machine-learning models, including Gaussian process regression. Ultimately, we find that encoding the functional form of the theory performs best followed by an encoding of the numeric values of the theory.

SUBMITTER: Audus DJ

PROVIDER: S-EPMC9912311 | biostudies-literature | 2022 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Leveraging Theory for Enhanced Machine Learning.

Audus Debra J DJ McDannald Austin A DeCost Brian B

ACS macro letters 20220826 9

The application of machine learning to the materials domain has traditionally struggled with two major challenges: a lack of large, curated data sets and the need to understand the physics behind the machine-learning prediction. The former problem is particularly acute in the polymers domain. Here we aim to simultaneously tackle these challenges through the incorporation of scientific knowledge, thus, providing improved predictions for smaller data sets, both under interpolation and extrapolatio ...[more]

PMID: 36018715

Similar Datasets

Project description:While many countries have registries of livestock farms, it can be challenging to obtain information on their primary production type. For example, for Swiss farms registered as keeping cattle, a distinction can only be made between milk-producing and non-milk-producing farms. The Swiss cattle industry consists of beef and dairy farms, with a strong predominance of small to medium-sized farms. A better differentiation of cattle production types would be beneficial for the planning and evaluation of surveillance programmes for cattle diseases and for the benchmarking antibiotic consumption. The aim of this study was to outline cattle production types of interest and to allow the classification of Swiss cattle farms according to production type in order to optimize surveillance. We collaborated with experts to define the five primary cattle production types: calf fattening, dairy cattle, cattle fattening, rearing cattle and suckler cows. In collaboration with the cantonal Veterinary Offices, we collected production types from 618 reference farms across 14 cantons and defined a total of 24 features by combining information from three national databases. Using farm-level data on milk production, age and sex distribution, cattle breeds, calving, births, slaughter, animal movements and antibiotic use, we trained three different machine learning models capable of classifying the five production types. Among these models, the Random Forest model demonstrated the highest level of performance, achieving an accuracy of 0.914 (95% CI: 0.890, 0.938) and an F1-Score of 0.879 (95% CI: 0.841, 0.913). In conclusion, together with experts, we have outlined five primary production types on cattle farms in Switzerland and developed a model that allows a reproducible, year-to-year classification of cattle farms using national datasets. Our flexible methodology could be adapted to other countries and datasets, enabling veterinary authorities to conduct more efficient and targeted disease surveillance in the future.

Dataset Information

Leveraging Theory for Enhanced Machine Learning.

Publications

Leveraging Theory for Enhanced Machine Learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets