A confidence-convergence decision layer for reliability-aware sequential recognition under limited observations

Description

This dataset supports the manuscript “A confidence-convergence decision layer for reliability-aware sequential recognition under limited observations”. It contains the figure source data tables, sensitivity-test data, publication-ready figure files, Python scripts for figure generation and data auditing, and the Supplementary Information files. The source data are provided as CSV files and are mapped to the corresponding main figures, including overall performance, coverage-risk behaviour, observation-window effects, confidence evolution, ablation analysis, and robustness analyses. The included scripts allow regeneration of the main figures and verification of the figure source data, supporting transparency and reproducibility of the reported results.

Steps to reproduce

1. Download and extract the dataset package. 2. Install the required Python dependencies listed in `requirements.txt`, or install NumPy, pandas, and Matplotlib in a Python 3 environment. 3. To audit the figure source data, run: `python scripts/audit_figure_data.py` 4. To regenerate the main figures from the source data, run: `python scripts/make_figures.py` 5. Compare the regenerated figures with the files provided in the `figures/` directory. The CSV files in `figure_source_data/` provide the source data for the corresponding main figures and sensitivity tests.

Institutions

• Wuhan University of Technology

  Hubei, Wuhan

Categories

Licence