NextFin news, An international team of researchers from institutions including INAF (Italy), the University of Parma (Italy), and the University of Waterloo (Canada) announced on Thursday, September 18, 2025, the development of Effort.jl, a novel AI software that simulates cosmic physics models on standard laptops with accuracy comparable to supercomputers.
Effort.jl emulates the Effective Field Theory of Large-Scale Structure (EFTofLSS), a complex theoretical model used to describe the vast cosmic web structure of the Universe. Traditionally, running such simulations required extensive time and computational power on supercomputers.
The breakthrough was published in the Journal of Cosmology and Astroparticle Physics (JCAP) and demonstrates that Effort.jl can deliver results with the same or even finer detail than the original models, but in just minutes on a typical laptop. This is achieved by combining neural networks with embedded physical knowledge, allowing the software to learn model responses efficiently and reduce the need for extensive training data.
Marco Bonici, a researcher at the University of Waterloo and lead author of the study, explained that the emulator uses gradients—measures of how predictions change with small parameter tweaks—to enhance learning efficiency. This approach drastically cuts computation time and resource requirements, enabling simulations that were once only feasible on supercomputers to be performed on everyday computing devices.
The team validated Effort.jl extensively using both simulated and real astronomical data, confirming its accuracy aligns closely with the original EFTofLSS model. In some cases, Effort.jl even allowed inclusion of analysis components that had to be omitted in traditional models due to computational constraints.
This advancement is particularly timely given the exponential growth of astronomical data from ongoing and upcoming surveys such as the Dark Energy Spectroscopic Instrument (DESI) and the Euclid mission. Effort.jl is expected to become a valuable tool for researchers analyzing these large datasets, facilitating faster and more accessible cosmic structure studies.
The research highlights the potential of AI-driven emulators to transform computational astrophysics by making high-fidelity cosmic simulations more accessible and efficient, thereby accelerating scientific discovery in understanding the Universe's large-scale structure.
Explore more exclusive insights at nextfin.ai.
