sahilugale/matchedfiltering

An interactive demonstration of matched filtering with some examples using real LIGO data and your computer's audio.

Quick start

Click the Binder button above to open a live version of the main notebook run in the cloud.

Alternatively, you can run the notebook on your own computer if you have anaconda installed. Just run these commands:

conda env create moble/gw_matched_filtering_demo
conda activate gw_matched_filtering_demo
gw_matched_filtering_demo

This will download and install all the necessary files, then open the notebook in your browser.

Details

This package is designed to be a simple demonstration of the principles of matched filtering. It uses the analogy of LIGO as a microphone to explain the basic ideas, using a microphone attached to the computer to study data as a function of time, noise sources, and real signals, as well as headphones or a speaker to play back those signals and others. Examples are given where a signal is buried in the noise and extracted using matched filtering. Real LIGO data and accurate gravitational waveforms are also included with this code, and used for further examples to complete the analogy. The concepts introduced here can be applied far more widely in all areas of data analysis.

Fourier transforms are introduced, starting with a simple example of a pure tone (which can be played by the computer), and progressing to Fourier transforms of noise and gravitational-wave signals. The matched filter is then introduced by progressively building the formula with simple explanations for each term. Discussion of these concepts is interwoven with practice using them.

The material is presented as a Jupyter notebook — which is an interactive python session, and includes text explaining the concepts and code. This allows the explanations to be included (with LaTeX equations) right among the code, and all in a live, interactive python session. No familiarity with python is necessary for the student, though the setup may require some basic skills.

To run the code

If you are familiar with python packaging, you can probably figure out how to run this on your own. Note that the required packages include ipython, jupyter, notebook, scipy, matplotlib, ipywidgets, and widgetsnbextension.

It is much simpler to just use the anaconda python ecosystem. Once anaconda is installed, just run the following at the command prompt:

conda env create moble/gw_matched_filtering_demo
conda activate gw_matched_filtering_demo
gw_matched_filtering_demo

This will install all the requirements into a new conda environment, switch to that environment, then download and run the notebook.

Notes for classroom use

There are three reasonable ways to deliver this demonstration to students: as a presentation, individually on the students' personal computers, and together in a computer lab.

Most likely, the presentation option is the least useful to students. Most students benefit enormously from being able to interact with the notebook personally. They will be more interested, able to read along at their own pace, and play with the parameters. If this is just not possible, it would be best to go slowly and ask lots of questions of the students, possibly allowing one student to actually run the commands while the teacher engages from off to the side.

A preferable option may be having the students download and run the code themselves. The only caveat here is that the students will need to install the dependencies. With anaconda, this is not a problem. Assuming the students can run it, there are questions included in the notebook. Their answers could be turned in as a homework assignment, or a quiz given on the material to ensure that students actually go through the notebook.

If this will be presented together in a computer lab, it is best if things are set up as much as possible on each computer beforehand. The computers need to be using different accounts (with home directories not on a shared file system), or ipython will get screwed up and run into errors.

Cool video

A student in one workshop pointed out this TED Talk by the astronomer Wanda Díaz-Merced who lost her sight, and now interacts with data by turning it into sound. Though she uses very different techniques from the ones we use for LIGO data, this is a very powerful example of the importance of exploring data in different ways.