Automatic Detection of Martian bow shock crossings using data of the Mars Express mission: A Deep Learning Approach to see more chek this link : https://pnst-2022.sciencesconf.org/399962.
We investigate to automatically detect the Martian bow shock crossings using the data of the Mars Express mission provided by CDPP-AMDA.
Using a Multilayer Perceptron Neural Network, we provide an automatic classifier to predict the
Martian bow shock crossings. A published catalog with around 11800 bow shock
crossings has been used for labeling the data [1]. The challenging task was to deal with
the unbalanced data, indeed, in our dataset, we have unequal distribution of classes:
shocks and no shocks. Classification of unbalanced data is a difficult task because
there are so few samples (shocks) to learn from. To tackle this problem is to penalize
the mis-classification made by the minority class by setting a higher class weight and at
the same time reducing weight for the majority class.
[1] B. E. S. Hall et al. “Annual variations in the Martian bow shock location as observed by the Mars Express mission”. In: Journal of
Geophysical Research: Space Physics 121.11 (2016), pp. 11, 474–11, 494
Example: python -m run_predictions_orchestra . --start 2008-07-03T00:00:00 --stop 2008-07-05T00:00:00
Zoom in between 2008-07-04T05:27:34 and 2008-07-04T08:40:20 =>
This repository contains a python script (run_predictions_orchestra.py) which should be executed in a virtual environment in order to make prediction of Martian Bow Shock crossings associated with a given start and stop times.
To do predictions you need to follow these steps: 👏- you should have python >= 3.10 already installed
- activate the virtual environment = venv\Scripts\activate
- install the requirements = pip --no-cache-dir install -r path/../requirements.txt
- run the script with the following arguments a destination folder path to store results, a start time start and a stop time stop => python -m run_predictions_orchestra . --start 2008-07-03 --stop 2008-07-06
👏