TY - GEN
T1 - A comprehensive analysis of machine learning algorithms for noisy image classification in astronomy
AU - Ntolkeras, George
AU - Moysis, Lazaros
AU - Lawnik, Marcin
AU - Maraslidis, George
AU - Choumpaev, Alexandra
AU - Manakitsa, Nikoleta
AU - Asimopoulos, Dimitrios Christos
AU - Fragulis, George F.
N1 - Publisher Copyright:
© 2026 American Institute of Physics Inc.. All rights reserved.
PY - 2026/6/16
Y1 - 2026/6/16
N2 - In this research, a systematic review of machine learning methods for noisy image classification in astronomy is provided. It is aimed at reviewing the developments and issues with different types of machine learning algorithms including supervised and unsupervised ones. This work also offers a detailed comparison of both deep learning structures including CNNs and Autoencoders and typical machine learning algorithms including Decision Trees, Random Forests, Support Vector Machines, and K-nearest neighbors. These algorithms and some more effective and complex versions of them are implemented in the Python environment and examined on the same dataset. The dataset contains noisy images collected from the solar system planets. The test results are compared, and the strong and weak points of each algorithm are identified and discussed. Finally, future possibilities for the development of similar approaches to increase the efficiency and reliability of astronomical image classification are considered.
AB - In this research, a systematic review of machine learning methods for noisy image classification in astronomy is provided. It is aimed at reviewing the developments and issues with different types of machine learning algorithms including supervised and unsupervised ones. This work also offers a detailed comparison of both deep learning structures including CNNs and Autoencoders and typical machine learning algorithms including Decision Trees, Random Forests, Support Vector Machines, and K-nearest neighbors. These algorithms and some more effective and complex versions of them are implemented in the Python environment and examined on the same dataset. The dataset contains noisy images collected from the solar system planets. The test results are compared, and the strong and weak points of each algorithm are identified and discussed. Finally, future possibilities for the development of similar approaches to increase the efficiency and reliability of astronomical image classification are considered.
KW - Astronomical Image Processing
KW - Deep Learning Algorithms
KW - Machine Learning in Astronomy
KW - Noisy Image Classification
KW - Supervised and Unsupervised Learning
UR - https://www.scopus.com/pages/publications/105042961270
U2 - 10.1063/5.0327695
DO - 10.1063/5.0327695
M3 - Conference contribution
AN - SCOPUS:105042961270
T3 - AIP Conference Proceedings
BT - AIP Conference Proceedings
A2 - Roy, Debopriyo
A2 - Fragulis, George
PB - American Institute of Physics
T2 - International Conference Series: Technology and Smart Society, ETLTC 2025
Y2 - 20 January 2025 through 26 January 2025
ER -