# spclustering: Super Paramagnetic Clustering Wrapper
This is a Python Ctypes wrapper for the original SPC algorithm (available in [Eytan Domany repository](https://github.com/eytandomany/SPC)). The SPC code was edited to work as a shared library, reduce disk usage and improve speed and interface with Python.
## How to install spclustering
```bash
pip install spclustering
```
## How to build spclustering
The original code requires gcc and makefile. For Windows I recommend [TDM-GCC](https://jmeubank.github.io/tdm-gcc/).
## How to use spclustering
```python
from spclustering import SPC
import matplotlib.pyplot as plt
import numpy as np
rng = np.random.RandomState(0)
cl1 = rng.multivariate_normal([8, 8], [[4,0],[0,3]], size=400)
cl2 = rng.multivariate_normal([0,0], [[3,0],[0,2]], size=1000)
data = np.concatenate([cl1,cl2])
clustering = SPC(mintemp=0.02,maxtemp=0.04)
results = clustering.run(data)
#select a temperature (similar to select a distance in a dendogram)
temp = 1
plt.plot(*data[results[temp,:]==0,:].T,'.r')
plt.plot(*data[results[temp,:]==1,:].T,'.b')
plt.plot(*data[results[temp,:]>1,:].T,'.k')
plot.show()
```
## Limitations and Changes
- It runs with a data matrix [npoints, ndims] as input. Running the clustering from a distance matrix is not implemented.
- Only the main parameters and flags are implemented.
- Tmax (stop temperature) is not included in the temperature array (similar to Python ```range```).
- Included *WriteEdges* boolean parameter (defaul False). If True, the \*.edges/\*.mst/\*.K file is created as in the original SPC. Added aswell *PrintParam* parameter in the legacy SW.c code.
- Reduces/removes console ouputs.
## Original SPC Citation
Blatt, M., Wiseman, S., & Domany, E. (1996). Superparamagnetic clustering of data. Physical review letters, 76(18), 3251.
```bibtex
@article{spc,
title={Superparamagnetic clustering of data},
author={Blatt, Marcelo and Wiseman, Shai and Domany, Eytan},
journal={Physical review letters},
volume={76},
number={18},
pages={3251},
year={1996},
publisher={APS}
}
```
Raw data
{
"_id": null,
"home_page": "https://github.com/ferchaure/SPC/",
"name": "spclustering",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "clustering",
"author": "Fernando Julian Chaure",
"author_email": "fchaure@fi.uba.ar",
"download_url": "",
"platform": null,
"description": "# spclustering: Super Paramagnetic Clustering Wrapper\n\nThis is a Python Ctypes wrapper for the original SPC algorithm (available in [Eytan Domany repository](https://github.com/eytandomany/SPC)). The SPC code was edited to work as a shared library, reduce disk usage and improve speed and interface with Python.\n\n## How to install spclustering\n\n```bash\npip install spclustering\n```\n\n## How to build spclustering\n\nThe original code requires gcc and makefile. For Windows I recommend [TDM-GCC](https://jmeubank.github.io/tdm-gcc/).\n\n\n## How to use spclustering\n\n```python\nfrom spclustering import SPC\nimport matplotlib.pyplot as plt\nimport numpy as np\n\nrng = np.random.RandomState(0)\n\ncl1 = rng.multivariate_normal([8, 8], [[4,0],[0,3]], size=400)\ncl2 = rng.multivariate_normal([0,0], [[3,0],[0,2]], size=1000)\ndata = np.concatenate([cl1,cl2])\n\nclustering = SPC(mintemp=0.02,maxtemp=0.04)\nresults = clustering.run(data)\n\n#select a temperature (similar to select a distance in a dendogram)\ntemp = 1\n\nplt.plot(*data[results[temp,:]==0,:].T,'.r')\nplt.plot(*data[results[temp,:]==1,:].T,'.b')\nplt.plot(*data[results[temp,:]>1,:].T,'.k')\n\nplot.show()\n\n```\n\n## Limitations and Changes\n\n- It runs with a data matrix [npoints, ndims] as input. Running the clustering from a distance matrix is not implemented.\n\n- Only the main parameters and flags are implemented.\n\n- Tmax (stop temperature) is not included in the temperature array (similar to Python ```range```). \n\n- Included *WriteEdges* boolean parameter (defaul False). If True, the \\*.edges/\\*.mst/\\*.K file is created as in the original SPC. Added aswell *PrintParam* parameter in the legacy SW.c code.\n\n- Reduces/removes console ouputs.\n\n## Original SPC Citation\n\nBlatt, M., Wiseman, S., & Domany, E. (1996). Superparamagnetic clustering of data. Physical review letters, 76(18), 3251.\n\n```bibtex\n@article{spc,\n title={Superparamagnetic clustering of data},\n author={Blatt, Marcelo and Wiseman, Shai and Domany, Eytan},\n journal={Physical review letters},\n volume={76},\n number={18},\n pages={3251},\n year={1996},\n publisher={APS}\n}\n```\n",
"bugtrack_url": null,
"license": "",
"summary": "Python Wrapper for Superparamagnetic Clustering",
"version": "0.0.3",
"split_keywords": [
"clustering"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "011486a421eb4d34b6c88d7a63589680d9d9afe6962704f62d2afb116e66a677",
"md5": "1aedcd928c03bccb3381f6544fbfa829",
"sha256": "b947b7e2b18cd20203af948a52b63360ee2d7b1cf7b4251a70c118a2e40cc674"
},
"downloads": -1,
"filename": "spclustering-0.0.3-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"has_sig": false,
"md5_digest": "1aedcd928c03bccb3381f6544fbfa829",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 63755,
"upload_time": "2023-01-07T21:53:26",
"upload_time_iso_8601": "2023-01-07T21:53:26.655597Z",
"url": "https://files.pythonhosted.org/packages/01/14/86a421eb4d34b6c88d7a63589680d9d9afe6962704f62d2afb116e66a677/spclustering-0.0.3-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-01-07 21:53:26",
"github": true,
"gitlab": false,
"bitbucket": false,
"github_user": "ferchaure",
"github_project": "SPC",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "spclustering"
}
JSON
