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 Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Exploratory landscape analysis (ELA) supports supervised learning approachesfor automated algorithm selection and configuration by providing sets offeatures that quantify the most relevant characteristics of the optimizationproblem at hand. In black-box optimization, where an explicit problemrepresentation is not available, the feature values need to be approximatedfrom a small number of sample points. In practice, uniformly sampled randompoint sets and Latin hypercube constructions are commonly used samplingstrategies. In this work, we analyze how the sampling method and the samplesize influence the quality of the feature value approximations and how thisquality impacts the accuracy of a standard classification task. While, notunexpectedly, increasing the number of sample points gives more robustestimates for the feature values, to our surprise we find that the featurevalue approximations for different sampling strategies do not converge to thesame value. This implies that approximated feature values cannot be interpretedindependently of the underlying sampling strategy. As our classificationexperiments show, this also implies that the feature approximations used fortraining a classifier must stem from the same sampling strategy as those usedfor the actual classification tasks. As a side result we show that classifierstrained with feature values approximated by Sobol' sequences achieve higheraccuracy than any of the standard sampling techniques. This may indicateimprovement potential for ELA-trained machine learning models.

Exploratory Landscape Analysis is Strongly Sensitive to the Sampling Strategy