C-SAM (Spectra analyzer machine for classification) and R-SAM (Spectra analyzer machine for redshift): An approach to spectral convolutional neuronal network

Spectral classification and redshift measurement present significant challenges in the field of contemporary astronomy. The complexity lies in the processing of data generated by cutting-edge telescopes, where current programs exhibit a pronounced reliance on outdated machine language. This limitation imposes a substantial barrier to the efficiency with which the vast volumes of data generated by innovative spectroscopic technologies can be analyzed. This challenge is magnified when dealing with extensive d atasets that require rapid and precise analysis. Therefore, it is crucial to initiate a quest for solutions to address these issues. While there is limited information regarding advances in the training of convolutional neural networks specifically designed for astronomy, notable progress has been observed in their application in other domains, such as the analysis of extensive datasets, the resolution of regression problems, and image processing. The potential for successfully applying these techniques to spectroscopy is promising. A prominent example can be found in the DESI project (Dark Energy Spectroscopic Instrument), which has implemented deep learning frameworks for the classification of galaxies, quasar objects (QSO), and Milky Way survey (MWS) objects. However, despite its potential, the magnitude of the collected data raises concerns about potential biases in object classification due to the inherent complexity of the task, which must be validated by other networks. This references the purpose of neural network SAM, where the primary challenge is to develop an effective classification system capable of accurately identifying the spectral class of millions of data points through the training of convolutional neural networks. As a result of this effort, the SAM neural network was successfully implemented, achieving impressive accuracy in classification and redshift measurement for the three spectral classes, given training with 200,000 spectra obtained from the DESI Project. This solution demonstrates its effectiveness in addressing the complexities of spectral classification and redshift measurement in astronomy as part of a computational problem.