# Bibliography
[Documentation](../README.md) / Reference
These sources identify definitions, theorems, and algorithmic ideas. They are not
runtime dependencies. Cocycle's engineering conventions are stated separately in
the [mathematical specification](mathematics.md).
## ZC05
Afra Zomorodian and Gunnar Carlsson. **Computing Persistent Homology**.
Discrete & Computational Geometry 33, 249–274 (2005).
[Publisher / DOI](https://doi.org/10.1007/s00454-004-1146-y).
Background for persistence over fields and interval classification. The project's
original source review checked bibliographic information and the abstract; the
specific reduction and pairing rules use the full-text source B21 below.
## B21
Ulrich Bauer. **Ripser: efficient computation of Vietoris–Rips persistence
barcodes**. Journal of Applied and Computational Topology (2021).
[arXiv v2](https://arxiv.org/abs/1908.02518v2),
[full text](https://arxiv.org/html/1908.02518v2),
[DOI](https://doi.org/10.1007/s41468-021-00071-5).
Relevant locations: §2 for Rips and filtration ordering; §3.1, Proposition 3.1 and
Algorithm 1 for reduction and pairing; §§3.2–3.5 for clearing, cohomology, implicit
matrices, and apparent pairs; §4 for H0/union-find. Cocycle independently implements
these invariants in Rust. Its internal order must satisfy each optimization's
conditions; citing Ripser does not establish correctness automatically.
## CSO13
Frédéric Chazal, Vin de Silva and Steve Oudot. **Persistence stability for
geometric complexes**.
[arXiv v3 (2013)](https://arxiv.org/abs/1207.3885v3),
[full text](https://arxiv.org/html/1207.3885v3).
Theorem 2.3 connects interleavings and bottleneck stability. Theorem 5.2 bounds
Rips diagrams using Gromov–Hausdorff distance. The fixed-vertex perturbation
argument in Cocycle first derives filtration inclusions, then applies Theorem 2.3.
## A20
Nieves Atienza, Rocío González-Díaz and Manuel Soriano-Trigueros.
**On the stability of persistent entropy and new summary functions for
Topological Data Analysis**. Pattern Recognition 107, 107509 (2020).
[arXiv v7](https://arxiv.org/abs/1803.08304v7),
[full text](https://arxiv.org/html/1803.08304v7),
[DOI](https://doi.org/10.1016/j.patcog.2020.107509).
Definition 3.1 uses base-two logarithms; §3.4 discusses infinite intervals.
Cocycle's natural logarithms, finite-only selection, and empty-result `None` are
explicit project conventions. The paper's title is not an unconditional stability
guarantee for every entropy variant.
## Architecture references
| GUDHI C++ | [Pinned source study](../research/gudhi-cpp.md) | Multiple representations and specialized engines; not globally tied to Simplex tree |
| PHAT | [Paper, §2](https://www.geometrie.tugraz.at/kerber/kerber_papers/phat_jsc.pdf) | Separate boundary-column storage from reduction strategy |
| Ripser | [Repository](https://github.com/Ripser/ripser), B21 | Specialized Rips computations can use implicit representations |
| giotto-tda | [API modules](https://giotto-ai.github.io/gtda-docs/latest/modules/index.html) | TDA includes capabilities beyond a single persistence pipeline |
These are design references, not a statement that Cocycle implements their APIs.
## External validation tools
Current comparisons use [pinned GUDHI and upstream Ripser C++ sources](../../benches/native/sources.json)
through [native workers](../../benches/native/README.md). Source versions, adapter
changes and precision are recorded with each run.
[GUDHI](https://gudhi.inria.fr/python/latest/) and
[Ripser.py](https://ripser.scikit-tda.org/en/latest/reference/stubs/ripser.ripser.html)
are optional historical development references. Pin versions and record precision, input,
coefficient field, scale, truncation, and interval conventions for each comparison.
The optional wrapper tools pin GUDHI 3.13.0 and Ripser.py 0.6.14. Ripser.py is a
Python binding/fork, distinct from the upstream C++ CLI. Its installed dense
input path converts distances to float32; the comparison suite uses shared
float32-exact inputs while retaining Cocycle/GUDHI f64 arithmetic. The
[online source](https://ripser.scikit-tda.org/en/latest/_modules/ripser/ripser.html)
helps locate this behavior but may describe a different version. See the
[benchmark protocol](../../benches/README.md) and recorded artifacts for actual runs.
## RP2
Sonia Balagopalan. **Small Triangulations of Projective Spaces**.
[Author's conference slides](https://www.maths.tcd.ie/~hmigca-18/slides/SoniaCGA.pdf).
The displayed six-vertex triangulation has facets 123, 124, 135, 146, 156, 236,
245, 256, 345, 346. Tests subtract one from these labels and take barycentric
subdivision to obtain a flag complex. Independent modular boundary ranks verify
the field-sensitive fixture; the source identifies the triangulation.
## CJS15
Nicholas J. Cavanna, Mahmoodreza Jahanseir and Donald R. Sheehy.
**A Geometric Perspective on Sparse Filtrations**. CCCG 2015.
[Author-hosted full text](https://donsheehy.net/research/cavanna15geometric.pdf),
[arXiv](https://arxiv.org/abs/1506.03797).
Relevant locations: §2 for finite metrics embedded in the max norm, §3 for the
covering lemma, §4 Theorem 4 for sparse nerve approximation, and §5 for Rips
simplex birth and disappearance constraints. The parameter conversion and
edge-length convention used by Cocycle are recorded in the
[specification](mathematics.md#15-sparse-rips-approximation).
[GUDHI's C++ documentation](https://gudhi.inria.fr/doc/latest/group__rips__complex.html)
states its `(1, 1/(1-epsilon))` convention; its pinned
[Sparse_rips_complex.h](https://github.com/GUDHI/gudhi-devel/blob/cba915e3ab8e1f5b1fe26eb44b407285f7af4e78/src/Rips_complex/include/gudhi/Sparse_rips_complex.h)
defines the compared edge/blocker algorithm. Theorem statements do not certify
floating-point implementations; native agreement and mathematical guarantees
are separate forms of evidence.