Shijun Cheng, and Tariq Alkhalifah

 GitHub: https://github.com/DeepWave-KAUST/Meta-LRPINN-pub

A new method called Meta-LRPINN, designed to model wavefields efficiently for various frequencies and velocity conditions. Traditional methods struggle with high computation costs and adapting to different scenarios. Meta-LRPINN solves these problems by combining advanced techniques: it uses a mathematical trick (SVD) to simplify the model, adds a smart way to include frequency information, and learns to start training from a better point using meta-learning. The results show this approach is faster, more accurate, and better at handling complex cases compared to current methods.