Sign in to view source links and access this dataset
Description
Six candidate configurations of an Inception-ResNetV2 model, each with a different permutation of three attention mechanisms (SE, SA, ACmix) across three network blocks. The table reports ROC-AUC, PR-AUC, F1-score, sensitivity, specificity, accuracy, and DeLong P values for pair-wise ROC-AUC comparisons with an SE-SE-SE baseline model. Author Chao Zhang uploaded this 9.3 KB XLSX file to figshare on 2026-05-14.
Use Cases
Benchmarking attention mechanism permutations based on reported ROC-AUC and PR-AUC metrics.
Statistical comparison of model variants based on DeLong P values for ROC-AUC.
Analyzing the trade-offs between sensitivity and specificity across different model configurations.
Strengths
Includes statistical significance testing via DeLong P values for ROC-AUC comparisons.
Reports six distinct model configurations evaluated on multiple performance metrics (ROC-AUC, PR-AUC, F1-score, sensitivity, specificity, accuracy).
Limitations
Row count is unknown, which may limit suitability assessment.
Column-level documentation is absent; field semantics must be inferred after download.
Provenance
Source
figshare
Collection Method
Likely generated from model training and evaluation experiments.
Freshness
Last updated 2026-05-14 17:44:31; freshness should be verified.
Dataset is very small (9.3 KB), indicating limited scope.