Forward Stratigraphic Modelling of Jezero Crater on Mars using Generative Adversarial Networks

我的项目是研究生成对抗网络（GAN）在正演地层建模中的潜在用例，当与基于扩散的建模技术相结合时。单次GAN（SinGAN）的使用将快速提供扩散模型的不确定性量化。我正在研究无约束GAN提供3D地层信息的有效性。这可以与各种形式的条件数据相结合，例如地震数据、钻孔日志和多光谱卫星图像，目的是在短时间内生成地质上可行的多个地层模型。

My project is investigating the potential use cases of Generative Adversarial Networks (GANs) in forward stratigraphic modelling, when combined with diffusion-based modelling techniques. The use of one-shot GANs (SinGAN) will be to rapidly provide uncertaintyquantification of diffusion models. I am researching the effectiveness of unconstrained GANs to provide 3D stratigraphic information. This can be combined with various forms of conditioning data, such as seismic data, borehole logs, and multi-spectral satellite imagery, with the aim of generating multiple stratigraphic models that are geologically feasible, in a short period of time.