Awesome

Automated classification of HAT sources

Overview

The directory setup is a bit overwhelming at the moment, but the basics of getting things up and running are:

• Run setup.py (build/install) to install the Cython LSP algorithm
• Install any dependencies, including
  • PyVislc
  • numpy
  • matplotlib
  • scipy
  • scikit-learn
  • mpi4py
  • hashlib
  • passlib
  • psycopg2
  • MySQLdb
  • pyfits
  • shlex
  • tornado
  • paramiko
  • spice/PySPICE: PySPICE, CSPICE

Install

• Edit the settings.py file to suit your needs this is also quite messy at the moment and will be cleaned up
• Run cd utils; python setup.py install; if you don't have permissions, you can append --user to the end of that.
• If you're running this on a remote system, make sure you have X11 forwarding on. Otherwise matplotlib will crash everything. Sorry; kind of annoying.

Running

1. Run create_initial_labeled_hatids.py to label and organize GCVS-crossmatched sources.
2. Run update_model.py to generate a classifier based on the labeled sources
3. Run get_candidates.py to search a HAT field for possible RR Lyr (or another user-specified subtype of variable star)
4. Run label_candidates.py to be written to visualize and manually label the candidates
5. Repeat 2 - 4 as many times as necessary until no more new sources are found.

Tutorials

###Connecting to della via terminal: * Use SonicWALL Mobile Connect to connect to a VPN * From terminal, do: $ ssh della

###Setting up environment * Clone the rrlyrclassifiation project to home directory: git clone git@github.com:johnh2o2/rrlyrclassification.git * [optional] Clone the pyvislc project to home directory: git@github.com:johnh2o2/pyvislc.git * Install additional modules * Run python setup.py install --user * Create SCRATCH directory in appropriate place; modify settings file * Run create_initial_labeled_hatids.py * On local system: python label_candidates.py --tarfile candidates_1.tar --dont-save --visualize