Comparison of Convolutional Neural Networks and Random Forest Classifiers for Strong Gravitational Lens Identification

Abstract

Strong gravitational lenses have been instrumental in providing insight into various astronomical problems, including analyzing the dark matter distribution of the universe. Effective identification of these events is made possible through machine learning algorithms, with many recent studies focusing on neural networks. However, very few have investigated the tradeoffs between different algorithms besides neural networks for lens identification. Our paper compares a convolutional neural network (CNN) and a random forest classifier (RFC) to determine the benefits of each for this task. We find that while CNNs do achieve higher accuracy, using RFCs to supplement them could increase the effectiveness of such algorithms. As a result, models that utilize both side-by-side to make predictions may increase in accuracy. This should be explored by future research.