Our purpose in research is not to focus on algorithm improvements and fine-tuning, though we end up downing so for cleaner and efficient results, but as academicians we are tasked with research goals that may look initially unreasonable, but have the potential to ideas and directions in our attempt to discover the Earth using seismic data. In our group, such ideas fall into the following focus categories:
Our experience in developing solutions in waveform inversion led to many interesting results. From the optimal transport matching filter (in action below on the Chevron blind dataset) to automatic waveform inversion based salt flooding (a movie on revolving the salt body iterative as part of FWI) and much more, we have, as a group learned the possibilities in seismic data waveform inversion and many of the challenges. With our recent endeavor in machine learning algorithms (since 2016), we managed to carve out innovate applications of machine learning in waveform inversion including the ML-misfit, the ML-decent, the ML-prior, and much more.
A new paradigm using the optimal transport matching filter in FWI.