| |
 |
| Paper: |
Machine Learning of Galaxy Classification by their Images and Photometry |
| Volume: |
535, Astronomical Data Analysis Software and Systems XXXI |
| Page: |
103 |
| Authors: |
Vavilova, I. B.; Dobrycheva, D. V.; Khramtsov, V.; Vasylenko, M. Y.; Elyiv, A. A. |
| Abstract: |
We discuss image- and photometry-based approaches for classification of 316,031 SDSS galaxies at z<0.1 with machine learning (ML). Using photometry parameters, human labeling, and five classical ML methods, we obtained that Support Vector Machine gives the highest accuracy for the binary galaxy morphological classification (96.1 % for early type, 96.9% for late type). Using CNNs as a model for the image-based inference of detailed morphology we attained 93% accuracy, on average, to classify these galaxies into five visual classes (completely rounded, rounded in-between, smooth cigar-shaped, edge-on, spiral). We obtained that CNNs with adversarial image data augmentation improves classification of smaller and fainter SDSS galaxies with mr < 17.7. We also introduce our approach to measure distance moduli m-M to the galaxies (accuracy is 0.35m) with a regression ML technique. |
|
|
|
|
|
