Comunicazione
Deep-Learning--based search for galaxy-scale lenses in the galaxy cluster environments.
Angora G., Rosati P., Meneghetti M., Brescia M., Mercurio A., Grillo C., Bergamini P., Acebron A., Caminha G., Nonino M., Tortorelli L., Bazzanini L., Vanzella E.
In the current era of big data, the development of methods able to autonomously extract information from vast multi-dimensional datasets plays a pivotal role. I will present how Convolutional Neural Networks (CNNs) can be trained to select galaxy-galaxy strong-lenses (GGSLs) in galaxy clusters. These systems can be used to characterize the subhalo component of the cluster mass distribution and test CDM structure formation paradigms. Although CNNs have been used to identify GGSL in the field, I will present how these cutting-edge algorithms can be tuned to detect such systems in the dense environment of galaxy clusters. The networks have been trained with simulated GGSLs, where sources have been injected in real HST image cutouts exploiting high-precision cluster-lens models (CLASH and HFF). Observations completely drive the simulation process, preserving the complexity of real data and producing simulated events indistinguishable from the real ones. This approach is extended beyond HST to Euclid mock data and VST imaging. The best CNNs achieve a high purity-completeness level (88%--93%). This methodology is applied to search for GGSLs around 6000 cluster members in 50 clusters.