Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/jacob-ian/pulsarclassifier
A machine learning Pulsar Classifier based on the LOTAASClassifier tool created by Rob Lyon of the University of Manchester. LOTAASClassifier can be reached at: https://github.com/scienceguyrob/LOTAASClassifier.
https://github.com/jacob-ian/pulsarclassifier
Last synced: 9 days ago
JSON representation
A machine learning Pulsar Classifier based on the LOTAASClassifier tool created by Rob Lyon of the University of Manchester. LOTAASClassifier can be reached at: https://github.com/scienceguyrob/LOTAASClassifier.
- Host: GitHub
- URL: https://github.com/jacob-ian/pulsarclassifier
- Owner: jacob-ian
- License: gpl-3.0
- Created: 2020-05-21T12:32:27.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2020-10-13T22:10:22.000Z (about 4 years ago)
- Last Synced: 2024-11-06T07:29:49.937Z (about 2 months ago)
- Language: Java
- Homepage:
- Size: 22.1 MB
- Stars: 0
- Watchers: 1
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: readme.md
- License: LICENSE
Awesome Lists containing this project
README
# PulsarClassifier
## Description
PulsarClassifier is a Machine Learning classifier that determines whether an object detected through a radio telescope is a pulsar or not. It is derivative of the LOTAASClassifier tool created by Rob Lyon (https://github.com/scienceguyrob/LOTAASClassifier).
The intention of this software is to implement the changes to LOTAASClassifier Version 1.0 (LC1) as discussed in Tan et al. (2017) that were added to the unreleased LOTAASClassifier Version 2.0 (LC2).
LC2 Features Implemented:
* [x] Ensemble Classifier
* [ ] Separate Radio Frequency Interference (RFI) Classification
## Requirements
* Java 1.6 or later
* Feature extracted candidate data produced with [PulsarFeatureLab](https://github.com/scienceguyrob/PulsarFeatureLab).
## Usage
### Step 1 - *Download*:
*Option A:*
* Download this software by runnning: `git clone https://github.com/jacob-ian/PulsarClassifier.git`
* Navigate to the `target` directory.
*Option B:*
* Download the `pulsarclassifier-1.0-full.jar` file from the `target` directory in this repository.
### Step 2 - *Train the Ensemble Classifier*:
* Use `PulsarFeatureLab` with the appended flags `--arff --meta` to extract the machine learning features from known pulsar and non-pulsar candidates. **NOTE:** Ensure that the feature type extracted is consistent across the training dataset and the data to be classified otherwise the machine learning classifier will not work.
* Edit the outputted file to remove the appended `?` character from each line and replace them with a `1` if the candidate is a pulsar, or a `0` if it a non-pulsar.
* Run the command:
```
$ java -jar pulsarclassifier-1.0-full.jar -t [PATH TO TRAINING DATASET ARFF FILE] -m [DIRECTORY TO STORE CLASSIFIER MODELS]
```
### Step 3 - *Make Predictions with the Ensemble Classifier*:
* Use `PulsarFeatureLab` with the appended flags `--arff --meta` to extract the machine learning features from your set of pulsar candidates. Do not edit the outputted file.
* Run the command:
```
$ java -jar pulsarclassifier-v1.0-full.jar -p [PATH TO DATASET ARFF FILE TO CLASSIFY] -m [DIRECTORY OF STORED CLASSIFIER MODELS]
```
### (Optional) Step 4 - *Testing and Validation*:
* Use [PulsarValidator](https://github.com/jacob-ian/PulsarValidator.git) to compare the list of pulsars in the testing/validation dataset with the output of the classifier.
## Output
PulsarClassifier will output 10 files in the same directory as the inputted `[output].arff` file. There are two files: a `[output]_[classifier].positive` file and an `[output]_[classifier].negative` file for each machine learning classifier, with the `.positive` file containing classified pulsars and the `.negative` file containing classified non-pulsars.
The ensemble classifier checks the result of the 4 machine learning classifiers and if a candidate is classified as a pulsar by 3 separate classifiers, it is classified as a pulsar by the ensemble classifier.
## License
PulsarClassifier is produced under the GNU General Public License Version 3.
## References
Lyon, R. J., B. W. Stappers, S. Cooper, J. M. Brooke, and J. D. Knowles. 2016. "Fifty years of pulsar candidate selection: from simple filters to a new principled real-time classification approach." *Monthly Notices of the Royal Astronomical Society* 459, no. 1 (April): 1104-1123. https://doi.org/10.1093/mnras/stw656.
Tan, C. M., R. J. Lyon, B. W. Stappers, S. Cooper, J. W. T. Hessels, V. I. Kondratiev, D. Michilli, and S. Sanidas. 2017. "Ensemble candidate classification for the LOTAAS pulsar survey." *Monthly Notices of the Royal Astronomical Society* 474(4): 4571-4583. https://doi.org/10.1093/mnras/stx3047