Accelerated and Energy-Efficient Galaxy Detection: Integrating Deep Learning with Tensor Methods for Astronomical Imaging

Addressing the astronomical challenges posed by the interplay of data volume, AI sophistication, and energy consumption is crucial for the future of astronomy. As astronomical surveys continue to produce vast amounts of data, the computational and energy demands for galaxy classification have escalated, necessitating more efficient and sustainable approaches. This study presents a novel application of tensor factorization within the Faster R-CNN framework, resulting in the development of our model, T-Faster R-CNN, designed to enhance both the energy efficiency and computational performance of deep learning models used in galaxy classification. By integrating tensor factorization, our T-Faster R-CNN significantly reduces the model’s complexity, memory footprint, and CO2 emissions, while maintaining, and in some cases even improving, the accuracy of morphological classification. The effectiveness of this optimized model is validated using data from the Galaxy Zoo DECaLS, where it demonstrates substantial improvements in computational efficiency without compromising classification precision. Furthermore, this research incorporates green code principles, emphasizing reductions in energy consumption and environmental impact in computational astronomy. The T-Faster R-CNN model offers a resource-efficient, sustainable methodology for analyzing large-scale astronomical data, addressing the critical need for greener computational practices in the era of big data.