Scientific Computing · Cluster
Reproducible Research
Notebooks, environments, and provenance.
- 01 Execution Counts and the Hidden State of a Notebook Kernel Why the cells stored in a notebook file were not necessarily executed in the order the file shows, what the execution counter measures, and which parts of the state never reach the document.
- 02 What a Notebook File Stores and What It Omits The version 4 notebook schema keeps source, rendered output, and a little kernel metadata in one JSON document, and drops the environment, the inputs, and the process that produced it.
- 03 Reproducibility, Replicability, and the Boundaries Between Them The distinct claims implied by re-running an analysis, rebuilding it from raw data on new instruments, and getting the same conclusion from an independent study, and which evidence each one requires.
- 04 Lockfiles, Hash Pinning, and Reconstructed Environments How a resolved dependency graph is recorded with content hashes, what the installer guarantees on a rebuild, and where the reconstruction stops being faithful.
- 05 Container Images as Frozen Environments and Their Limits Layer digesting, reproducibility of the build itself, kernel sharing, and the hardware facts a container cannot pin down.
- 06 Seeds, Generator State, and Reproducible Random Streams Why a seed is not enough to reproduce a stochastic result, how generator state advances, and how parallel workers avoid splitting one stream into correlated pieces.
- 07 Content-Addressed Provenance Graphs for Derived Outputs Recording each step as a node keyed by the hash of its inputs, code, parameters, and environment, so that any output can be traced back to the exact bytes and commands that produced it.
- 08 Conditioning, Tolerances, and the Digits a Reproduction Can Demand How the condition number sets the width of the interval a computed result actually occupies, and how that interval determines which differences between two runs are meaningful.
- 09 Bit Rot, Link Rot, and Replaying an Analysis Years Later What degrades in a stored computation over time, which failures are recoverable, and what has to be captured at creation time to make a late replay possible at all.
- 10 Executing Provenance Records to Certify That a Result Still Reconstructs Turning a recorded dependency graph into a re-runnable pipeline that hashes outputs and compares them to stored digests, so a result carries continuous evidence that it is still reproducible.