Reproduce “DeepWaveRL: Self-Supervised Full Waveform Inversion via Reinforcement Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Peng Jin, Yinan Feng, Yinpeng Chen, Youzuo Lin, Sharon Huang
DeepWaveRL performs self-supervised seismic velocity-map inversion without velocity-map labels or simulator gradients. It samples candidate maps, simulates their seismic traces, and uses scalar data-misfit rewards to guide policy updates. On synthetic OpenFWI CurveVel-A, with test-time optimization, it approaches supervised InversionNet but scores worse on all three metrics: MAE 0.0527 versus 0.0409, RMSE 0.1012 versus 0.0944, and SSIM 0.8601 versus 0.8796.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.