Welcome to GravitySpy’s documentation!

Gravity Spy is an innovative citizen-science meets Machine Learning meets gravitational wave physics project. This repository is meant to faciliate the creation of new similar citizen science projects on Zooniverse

The module level docstrings attempt to follow the following format : Google Style Sphinx

Installing GravitySpy

The easiest method to install gravityspy is using pip directly from the GitHub repository:

Gravity Spy software has been tested on py35 and py27. Work is in progress to have unit tests verify gravity spy on py27 and py34 py35 and py36.

$ pip install git+https://github.com/Gravity-Spy/GravitySpy.git

For more details see Installation.

Publications

If you use Gravity Spy in your scientific publications or projects, we ask that you acknowlege our work by citing the publications that describe Gravity Spy.

How to classify unlabeled excess noise

One of the main product of this package is the command-line executable wscan, which takes an excess noise time makes an omega scan of the event and classifies the image.

To run an analysis:

$ wscan --inifile my-wini-config-file.ini --eventTime <eventTime> --outDir ./public_html/GravitySpy/Test/ --uniqueID --ID 123abc1234 --HDF5 --runML --pathToModel ../bin

where <eventtime> is the GPS time stamp assosciated with an Omicron trigger or other time of known excess noise, and ./my-wini-config-file.ini is the path of your configuration file. In the folder Production, is an example of an ini file.

For a full list of command-line argument and options, run

$ wscan --help

For more details see Running wscan on the command line.

How to train a new model

Another main product of this package is the command-line executable trainmodel, which serves as an easy wrapper function for training a new Convolutional Neural Net Model (CNN).

To train a model:

$ THEANO_FLAGS=mode=FAST_RUN,device=cuda,floatX=float32 trainmodel --path-to-trainingset <path-to-trainingset> --number-of-classes <number-of-classes>

where <path-to-trainingset> is a folder that has the structure “class”/”sampes” and <number-of-classes> is how many different classes there are.

For a full list of command-line argument and options, run

$ trainmodel --help

For more details see Running trainmodel on the command line.

The many databases of GravitySpy

For more details see The many DBs of Gravity Spy

Indices and tables