Skip to content
Publications

Dataset · Zenodo · September 12, 2026

Worst-case bounds on classifier comparisons under image-group link loss: reproducibility archive

Bojian Guo, Xin Li, Nazgul Ismanovna Temirbaeva, A.Dzh. Kartanova

Abstract

Reproducibility archive for the manuscript 'Worst-case bounds on classifier comparisons under image-group link loss'. It contains the exact 1,797-image public benchmark subset used in the study, labels, deterministic transformations, split assignments, selected parameters, fixed predictions, sensitivity results, attainable-bound witnesses, figures, checksums, and the code required to audit the reported calculations. The released image representations originate from the cited UCI Optical Recognition of Handwritten Digits dataset (DOI: 10.24432/C50P49; CC BY 4.0). Shifted and blurred views are explicitly defined computational treatments, not newly collected observations. Original code is licensed under MIT; original derived outputs and documentation are licensed under CC BY 4.0.

Machine Learning and Data ClassificationAdvanced Neural Network ApplicationsData Mining and Machine Learning Applicationsclassifier comparisonworst-case boundsreproducibility

More by Asel Kartanova