Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Recurso digital |
| Language: | English |
| Published: |
Zenodo
2026
|
| Subjects: | |
| Online Access: | https://doi.org/10.5281/zenodo.19666286 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>This artifact accompanies "Effective Structural Entropy and Semantic Robustness in Error-Tolerant Hidden Markov Models" and provides the complete analysis pipeline for the paper's quantitative study. It contains the Rust simulation engine, all Python analysis scripts, 25 pre-generated Parquet data files (10.5 M perturbed evaluation pairs), and the paper PDF and LaTeX source.<br><br>Reproducibility. All quantitative results in the paper are reproducible, including the three paper figures, all four tables, and all 1307 computed statistical values (correlation coefficients, effect sizes, regression slopes, failure rates, and composite index values). A consistency script verifies that every recomputed value appears in the paper and analysis documentation, and that every numeric claim in the paper is backed by a registry entry.<br><br>Platform. The bundled Docker image is built for linux/amd64. It runs natively on x86_64 Linux and x86_64 macOS, and under Docker Desktop's built-in cross-platform emulation on arm64 hosts (Apple Silicon, ARM Linux). A native arm64 image is not provided because a Rust dependency has a compilation bug on that target. The artifact was prepared and tested on macOS 15 Apple Silicon M3. Docker is the only host requirement.<br><br>Evaluation time (measured on Apple Silicon M3 under emulation; x86_64 will be faster). The artifact provides five graduated tiers. A smoke test (Tier 1) runs in about 4 min. Tiers 1 and 2 together (about 5 min) verify all paper figures and numeric consistency from the bundled data. Full recomputation of all statistical values takes about 10 min (Tier 3). All paper and supplementary figures can be regenerated in about 3 min (Tier 4). A full end-to-end re-run from source, covering Rust simulation, statistical recomputation, and figure regeneration, takes about 1.5 h (Tier 5).<br><br>Network access. Not required. The artifact is fully self-contained and all dependencies are bundled in the Docker image. Network access is only needed if rebuilding the Docker image from source. The Dockerfile and all dependency lock files are included for this purpose.<br><br>Future-proofness. All Rust and Python dependencies are pinned and baked into the bundled Docker image. No packages are downloaded at runtime. Base images reference stable versioned tags (rust:1.89-bookworm, python:3.12-slim-bookworm).<br><br>Special hardware and licenses. None. Code is released under MIT. Data and paper are under CC-BY 4.0.</p>