Scaling Laws for Galaxy Images

arXiv:2404.02973v1 Announce Type: new
Abstract: We present the first systematic investigation of supervised scaling laws outside of an ImageNet-like context - on images of galaxies. We use 840k galaxy images and over 100M annotations by Galaxy Zoo volunteers, comparable in scale to Imagenet-1K. We find that adding annotated galaxy images provides a power law improvement in performance across all architectures and all tasks, while adding trainable parameters is effective only for some (typically more subjectively challenging) tasks. We then compare …

abstract annotations arxiv astro-ph.ga context cs.cv galaxy imagenet images improvement investigation law laws power scale scaling type