SNN Tagging
A tool to recover golobular clusters with chemical abundances and radial velocities.Documentation
Obtaining/Installation
How to install.
Currently SNN_Tagging is available as a github project. Clone using:
git clone https://github.com/DOnghiaGroup/SNN_Tagging
Requirements.
The following packages are required to use SNN_Tagging. You can install them with pip or conda (or your prefered package manager).
pip install --user --upgrade astropy
pip install --user --upgrade sklearn
pip install --user --upgrade scipy
pip install --user --upgrade numpy
pip install --user --upgrade matplotlib
In addition, most of the tools are currently implemented as Jupyter notebooks. Make sure you have a recent version of Jupyter installed.
Quickstart
From the command line, move into the SNN_Tagging folder that you created above.
cd SNN_Tagging
And start up a Jupyter notebook server.
Jupyter notebook
Follow along in the notebook ‘SNN_data_analysis_example.ipynb’
About
SNN stands for Shared Nearest Neighbors. This is the technique we use to combine disparate information about stars. In this case information about the star’s radial velocity and chemistry.
See our paper here: coming soon
Running
Note: The full analysis currently takes ~20 hours to run on a moderatly fast machine. If you are curious about how it works, we encourage you to look first at the quickstart guide above.
For this analysis, we assume that you have the stellar parameter catalog derived from The Cannon. If you are a SDSS-IV member you can find this at the SDSS wiki (login required). We specifically use file “results-unregularized-matched.fits.gz”.
If you want to use a different stellar parameter catalogs, you will need to modify the variable FILENAME. Some of the analysis below, may not work on all datasets. If you are interested in applying this method to other data, please let us know and we will be happy help.
The data analysis part is run with SNN_data_analysis.py. Before running the file, values need to be assigned to the following variables.
EPS = 0.35
MIN_SAMPLES = 8
N_CHEM = 500
N_RV = 300
N_CUT = 1
N_CHEM is the number of nearest neighbors considered in the chemical space. N_RV is the number of nearest neighbors considered in the velocity space. N_CUT is the minimum amount of neighbors that a star considered for clustering should have. EPS and MIN_SAMPLES are parameters for DBSCAN clustering algorithm.
| The data analysis is currently only run on stars with | b | < 20 degrees. |
After the data analysis is finished, the following files are created as output.
ap_table_halo_nn
SNN_DBSCAN_labels.p
SNN_distance_matrix.p
ap_table_halo_nn contains the original dataset used for clustering. SNN_DBSCAN_labels has the clustering results from DBSCAN. SNN_distance_matrix is the distance matrix used for DBSCAN clustering.