Feasibility of probing the correlation function via deep learning


  • Denny Lane B. Sombillo National Institute of Physics, University of the Philippines Diliman
  • Neris I. Sombillo Department of Physics, Ateneo de Manila University


We investigated the feasibility of utilizing deep learning techniques to extract information regarding the range of source size and the characteristics of hadronic interactions from correlation functions. Initially, we conducted a preliminary analysis by training six distinct deep neural network models using the Lednický -Lyuboshits model. The objective was to identify the range of source sizes within the correlation function and determine the signs of scattering length and effective range for the corresponding scattering amplitude. To ensure effective learning for the classification task, the training dataset was structured in a curriculum fashion. Among the various neural networks considered, the optimal design comprised three hidden layers, each consisting of 250 nodes. This particular configuration yielded a final accuracy of approximately 97%. The outcomes of this study serve as a benchmark for evaluating more complex correlation functions beyond the simplified Lednický-Lyuboshits model.



Article ID



Poster Session C (Mathematical Physics, Optics, and Interdisciplinary Topics)



How to Cite

DLB Sombillo and NI Sombillo, Feasibility of probing the correlation function via deep learning, Proceedings of the Samahang Pisika ng Pilipinas 41, SPP-2023-PC-29 (2023). URL: https://proceedings.spp-online.org/article/view/SPP-2023-PC-29.