Team: 50
School: St. Thomas Aquinas
Area of Science: Physics and Astronomy
Interim: Problem Definition: Modern astronomical telescopes are looking deeper into the universe. The images that they produce reveal the existence of countless stars and galaxies. The seemingly infinite quantity of these astronomical objects in these images has become too much for humans to identify and characterize. Machine learning can provide the ability to automatically identify these astronomical objects.
Problem Solution: Using computer programming (Python) and machine learning, train a neural network to classify galaxies of interest that exist within telescope images.
It is important because computers can preprocess the image and indicate to the astronomer the existence of specific objects or phenomena of interest. This helps identify new astronomical discoveries.
Progress to Date: Currently, utilizing python's scikit-learn machine learning module to classify images of galaxies between being spiral or elliptical, the computer running on Intel i9-9900kf running all cores, took about an hour to train with a 1,000 image data set to complete the task with 78% accuracy. Further development will take place using python's TensorFlow deep learning module to gain a higher classification accuracy.
Expected Results: The primary goal of this supercomputing effort is to develop a machine learning model that is capable of classifying galaxies in the galaxy-zoo 2 dataset with high-accuracy, preferably above 95%. Classifications will focus on spiral and elliptical galaxies at first, but may include edge-on, ring, and irregular galaxies if time and the dataset will allow. The final program will provide a tk-GUI showing the galaxy image, its classification truth, and the machine learning model’s predicted classification.
Works Cited:
Briggs, J. R. (2013) . Python for Kids: A Playful introduction to programming. San Francisco: No
Starch Press Inc.
Broxton, M. J. (2010). Automated Classification of Galaxy Zoo images. (CS299, Stanford
University, 2010). Final Report.
Chen, J. (2020) . Galaxy morphology classification based on an improved deep convolutional
neural network. Journal of Physics.: Conf. Ser. 1549 042033.
Coccomini, D., Messina, N., Gennaro, C., & Falchi, F. (2021) . Generative adversarial networks
for astronomical images generation. (University of Pisa Italy, Nov 2021).
arXiv:2111.11578v1[cs.CV] 22 Nov2021.
Raschka, S., & Mirjalili, V. (2019) . Python Machine Learning Third Edition. Birmingham, UK:
Packt Publishing Ldt.
Liu, Y. H. (2020) . Python Machine Learning By Example Third Edition. Birmingham, UK: Packt
Publishing Ldt.
Walmsley, M., et al., (2019) . Galaxy Zoo: Probabilistic morphology through Bayesian CNNs and
Active Learning. (University of Oxford, Oct 2019). arXiv:1905.07424v2[astro-ph.GA]
4Oct2019.
Willett, K. W., et al., (2013) . Galaxy Zoo 2: detailed morphological classifications for 304 122
galaxies from the Sloan Digital Sky Survey. (University of Minnesota, Jun2013).
https://academic.oup.com/mnras/article/435/4/2835/1022913
17Jun2013.
Team Members: Catherine Sedillo
Mentor: James Sedillo
Sponsoring Teacher: Eric Vigil
Team Members:
Sponsoring Teacher: Eric Vigil