Live Pulse Finder

This repository contains a method for detecting transient astronomical events in realtime.

Installation

Requirements:

• A python3 (>3.4) installation with working pip3 and python3-venv.

1. Clone the repository.

git clone [email protected]:transientskp/lpf.git

2. Run Installation

Installation script is provided in activate.sh which you can run as follows:

cd lpf/
. activate.sh

Otherwise, check the requirements.txt file.

---

In the following, you have to setup a configuration file. Consider copying one of the ones provided in examples/ and editing it to your needs. We use examples/aartfaac12.yml in the guidelines.

Transient Simulation

To train the neural network for inference, we first build a dataset.

1. Create a parameter configuration file accustomed to your telescope. See the examples folder for inspiration.
2. Run the lpf/simulation/scripts/transients.py script with as argument the path to your configuration file. E.g.,

python lpf/simulation/scripts/transients.py examples/aartfaac12.yml

3. Once the simulation is finished, some example PNGs will be given in the output folder that you provided in the configuration file. Make sure they look satisfactory.

Noise Extraction (Optional)

This extracts background noise for the dynamic spectra. If skipped, you'll use Gaussian noise.

1. Specify correct parameters in the noise extraction section of your configuration file.
2. Run the noise extractor:

python lpf/simulation/scripts/extract_noise.py examples/aartfaac12.yml

Neural Network Training

1. Edit the neural network section of your configuration file to your needs.
2. Run

python lpf/_nn/scripts/train.py examples/aartfaac12.yml

3. Wait until it's converged.

Run LPF

1. Edit your configuration file to your needs.
2. Run

python lpf/main.py examples/aartfaac12.yml

3. The parameters of analyzed transients will be output to a .csv in the specified output folder. This can be opened for analysis. The .npy file in the output folder constains all the dynamic spectra.

Analyse Results

1. The .csv file with the inferred parameters is in the output folder you specified. You can use pandas to inspect it and filter it for interesting bursts. An example .ipynb file is given in lpf/analysis/result_analysis.ipynb.
2. Also: in the provided output folder a catalog video is saved to show the source-detection pipeline and an example of the estimated background and variability maps are saved.