Copyright (C) 2022 Parker Edwards
- Julia. Written and tested with version 1.4.1.
- Jupyter notebook
- HomologyInferenceWithWeakFeatureSize.jl.
- (Optional) If you would like to re-compute bottleneck correspondence solutions, Bertini, parallel version recommended.
- Clone this repository
- In a terminal, boot Julia, go to package mode with
]
. There are now two options - A - If you are not worried about e.g.
HomotopyContinuation.jl
in your main Julia environment, you can just runadd https://github.com/P-Edwards/HomologyInferenceWithWeakFeatureSize.jl
- B - If you are concerned about this. First run
activate .
. Then runadd https://github.com/P-Edwards/HomologyInferenceWithWeakFeatureSize.jl
. - (Optional) If you would like to run the
butterfly_lfs_sparse
example, alsoadd Distances, PersistenceDiagrams, ProgressBars, JLD, CSV, Measures, Plots, DataFrames
- B - With option B, a
Project.toml
file was created. Copy this into any directory with a notebook you would like to run yourself.
This repository contains data and scripts accompanying the manuscript Computing geometric feature sizes for algebraic manifolds. The directory structure is sorted by example. To recompute Julia-based filtering steps with precomputed solutions, install the first three requirements and:
- Start jupyter notebook via terminal with
jupyter notebook
- Navigate in the browser tab opened to a notebook file in one of the provided directories.
If you would like to check the algebraic computations, each directory for an example which uses Bertini includes a directory bertini_inputs
. Each leaf folder contains a file labelled input
, and you can run Bertini, say with 12 threads, via mpirun -np 12 bertini input
. Follow directions in the notebooks to filter these new solutions.
Some data files are too large to store on GitHub. They are instead available for download at https://notredame.box.com/v/feature-size-large-data. For each directory at that link, place the data in the "paper_data" folder for this repository's correspondingly named directory.
Algebraic computations for 3- and 4-bottlenecks for the examples in R^3, i.e. the quartic surface and torus-Clebsch intersection curve, are expensive. The 4-bottlenecks in particular can take a week or more to re-compute with Bertini.
This repository is distributed under GPLv3.