English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Open

Presents the pdf analysis as premium feature

PDF Analysis

Insights

Presents the summarisation as premium feature

Summarisation

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Machine learning techniques typically rely on large datasets to createaccurate classifiers. However, there are situations when data is scarce andexpensive to acquire. This is the case of studies that rely on state-of-the-artcomputational models which typically take days to run, thus hindering thepotential of machine learning tools. In this work, we present a novelclassifier that takes advantage of lower fidelity models and inexpensiveapproximations to predict the binary output of expensive computer simulations.We postulate an autoregressive model between the different levels of fidelitywith Gaussian process priors. We adopt a fully Bayesian treatment for thehyper-parameters and use Markov Chain Mont Carlo samplers. We take advantage ofthe probabilistic nature of the classifier to implement active learningstrategies. We also introduce a sparse approximation to enhance the ability ofthemulti-fidelity classifier to handle large datasets. We test thesemulti-fidelity classifiers against their single-fidelity counterpart withsynthetic data, showing a median computational cost reduction of 23% for atarget accuracy of 90%. In an application to cardiac electrophysiology, themulti-fidelity classifier achieves an F1 score, the harmonic mean of precisionand recall, of 99.6% compared to 74.1% of a single-fidelity classifier whenboth are trained with 50 samples. In general, our results show that themulti-fidelity classifiers outperform their single-fidelity counterpart interms of accuracy in all cases. We envision that this new tool will enableresearchers to study classification problems that would otherwise beprohibitively expensive. Source code is available athttps://github.com/fsahli/MFclass.

Multi-fidelity classification using Gaussian processes: accelerating the prediction of large-scale computational models