https://github.com/szcf-weiya/monotonesplines.jl
Monotone Cubic B-Splines with a Neural-Network Generator
https://github.com/szcf-weiya/monotonesplines.jl
b-splines monotone neural-network
Last synced: 3 months ago
JSON representation
Monotone Cubic B-Splines with a Neural-Network Generator
- Host: GitHub
- URL: https://github.com/szcf-weiya/monotonesplines.jl
- Owner: szcf-weiya
- License: mit
- Created: 2022-10-18T22:16:23.000Z (over 2 years ago)
- Default Branch: master
- Last Pushed: 2025-01-27T02:29:36.000Z (5 months ago)
- Last Synced: 2025-02-28T16:00:31.139Z (4 months ago)
- Topics: b-splines, monotone, neural-network
- Language: Julia
- Homepage: https://hohoweiya.xyz/MonotoneSplines.jl/
- Size: 43.3 MB
- Stars: 5
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# MonotoneSplines.jl
[](https://szcf-weiya.github.io/MonotoneSplines.jl/dev)
[](https://github.com/szcf-weiya/MonotoneSplines.jl/actions/workflows/ci.yml)
[](https://codecov.io/gh/szcf-weiya/MonotoneSplines.jl)
*MonotoneSplines.jl* is a Julia package for monotone splines, which impose a monotonicity constraint on the smoothing splines.
Check the following paper for more details.
> Lijun Wang, Xiaodan Fan, Huabai Li, and Jun S. Liu (2024). “Monotone Cubic B-Splines with a Neural-Network Generator.” Journal of Computational and Graphical Statistics. https://doi.org/10.1080/10618600.2024.2431070
```
@article{wangMonotoneCubicBSplines2023b,
title = {Monotone Cubic {{B}}-Splines with a Neural-Network Generator},
author = {Wang, Lijun and Fan, Xiaodan and Li, Huabai and Liu, Jun S.},
year= {2024},
journal = {Journal of Computational and Graphical Statistics},
date = {2024-09-28},
eprint = {2307.01748},
eprinttype = {arxiv},
eprintclass = {astro-ph, stat},
doi = {10.1080/10618600.2024.2431070},
url = {https://doi.org/10.1080/10618600.2024.2431070},
urldate = {2023-09-21},
pages = {in press},
keywords = {published}
}
```
## :hammer_and_wrench: Installation
:bangbang::bangbang: **R is required, but Python is optional.** :bangbang::bangbang:
The package offers two deep-learning backends for the Multi-Layer Perceptrons (MLP) generator:
- [Flux](https://github.com/FluxML/Flux.jl): pure Julia
- PyTorch: Call Python via [PyCall.jl](https://github.com/JuliaPy/PyCall.jl)
The statistical software R has offered several powerful packages on splines, such as `splines` and `fda`. We do not reinvent the wheel. Instead, we stand on the shoulders of the R giant by calling several basic core functions with the help of [RCall.jl](https://github.com/JuliaInterop/RCall.jl).
### :gear: use system R and Python
*MonotoneSplines.jl* is available at the General Registry, so you can easily install the package in the Julia session after typing `]`,
```julia
julia> ]
(@v1.8) pkg> add MonotoneSplines
```
By default, both `PyCall.jl` and `RCall.jl` would try to use the system Python and R, respectively (more details can be found in their repos).
### :ladder: standalone R and Python via Conda.jl
Another easy way is to install standalone `R` and `Python` via [Conda.jl](https://github.com/JuliaPy/Conda.jl) (no need to install `Conda.jl` explicitly) by specifying the following environmental variables before adding the package.
```julia
julia> ENV["PYTHON"]=""
julia> ENV["R_HOME"]="*"
julia> ]
(@v1.8) pkg> add MonotoneSplines
```
If you use the standalone R provided by Conda in Julia, the dependent R packages will be automatically installed during the building step.
## :books: Documentation
The documentation elaborates on the usage of the package via various simulation examples and an interesting astrophysics application.