Simulation-Based Inference for Weak Lensing: From Stage III to Stage IV Surveys

Simulation-based inference (SBI) and higher-order statistics have emerged as powerful tools for extracting significantly more cosmological information from large-scale structure probes. Their implementation, however, requires a genuinely multi-disciplinary approach, combining high-performance computing (e.g., large N-body simulations, parallel optimization, and data-intensive workflows), detailed physical and observational modeling (including baryonic feedback, intrinsic alignments, and survey systematics), and advanced statistical and machine-learning methods tailored to these frameworks (such as data compression and neural likelihood estimation). While these methods are relatively recent, they have now reached a level of maturity in the context of weak lensing analyses. In this talk, I will present the current state of the art, focusing on results from Stage III surveys based on DECam data, covering nearly 15,000 square degrees of the sky. I will then discuss the transition to Stage IV surveys, highlighting both the opportunities and the key challenges that must be addressed to fully realize the potential of SBI
in the next generation of cosmological analyses.

Zoom: https://dipc-org.zoom.us/j/98041357548
YouTube: https://youtube.com/live/3qAD-mI-30Q