A repository for the use of artificial neural networks to automatically classify images of the resonant argument of the M1:2 mean-motion resonance. The codes were developed for a Linux Ubuntu operating system, with gfortran, python3.8.5 and tensorflow installed. They may be portability issues for other platforms. In the main branch there is a UNIX script (script_analysis) that will automatically perform the analysis of images of resonant arguments for simulations in the 01 branch.
A sample of 99 images of resonant arguments and their labels, listed in the file all_m12_status_new, used to train the neural networks, are available in the branch ALL_PNG_varpic. The file all_m12_status_new reports the asteroid identification, the values of proper semi-major axis, eccentricity and sine of inclination, the asteroid absolute magnitude, and the status of the asteroid, 0 for asteroids in circulating orbits, 2 for asteroids on librating states, and 1 for objects alternating phases of libration and circulation. The images show the time evolution of the resonant argument for the M12 resonance, 2\lambda- {\lambda}_M+{\varpi}_M, where {\lambda} is the true anomaly, {\varpi} is the longitude of pericenter, and the suffix M identifies the planet Mars. A zip file containing the full database of 5700 images and labels used in Carruba et al. (2021), MNRAS, 504, 692, can be downloaded from:
https://drive.google.com/file/d/1RsDoMh8iMwZhD-fnkYSs9hiWmg96SZf0/view?usp=sharing.
Or, alternately, using curl:
curl -c ./cookie -s -L "https://drive.google.com/uc?export=download&id=1RsDoMh8iMwZhD-fnkYSs9hiWmg96SZf0" > /dev/null
curl -Lb ./cookie "https://drive.google.com/uc?export=download&confirm=`awk '/download/ {print $NF}' ./cookie`&id=1RsDoMh8iMwZhD-fnkYSs9hiWmg96SZf0" -o ALL_PNG_varpic.zip
And then unpacking the zip file. Several codes used for the numerical simulations and their analysis are available in the branch CODES. swift_bs.f is a Burlisch-Stoer integrator from the SWIFT package (Levison and Duncan 1994, Icarus 108, 18, available at https://www.boulder.swri.edu/~hal/swift.html). follow_all_filtered_plan_el.f will transform the outcome of the numerical simulation from a binary file to a ascii format. res_arg_m12.f will compute the resonant arguments for the M1:2 resonance for 50 test particles in the simulation (it needs a file saturn to be in the same directory). plot_rez_id.py will plot the time behavior of the resonant arguments. image_class.py will classify the images based on training obtained from images and labels at the image_class.py branch. Finally, an example of a simulation of 50 asteroids with all the necessary input and output files is available in the RUN_00/01 branch.
In order to run the model, users will only have to run the script, after changing its ṕermissions, using the following commands (the file script_analysis is available in the branch RUN_00):
chmod+x script_analysis
./script_analysis
Questions and comments can be addressed to mlasb2021@gmail.com.