Physics-Informed Neural Networks for Exploration Seismology
Seismic Series
Where deep learning meets the wave equation, and both come out stronger.
From "what is a neuron?" to confident, working expert in PINN methods for seismic inverse problems.
13 parts
92 sections
Free, browser-native
Table of contents
Every section is a working session: text, math, code, interactive widgets. Click any title to jump in.
Part 1: Part 0: Neural networks from absolute zero
- What is a neuron?
- Layers, depth, and universal approximation
- Activation functions and inductive bias
- Loss functions and the loss landscape
- Gradient descent by hand
- The chain rule and backpropagation
- Auto-differentiation as a computational graph
- Training loops and optimisers
- Spectral bias and why physics needs it fixed
- A geophysics primer for ML readers
Part 2: Part 1: The PINN formulation
Part 3: Part 2: Architectures for PINNs
Part 4: Part 3: Training pathologies and remedies
- Why most beginner PINNs do not converge
- The loss-balance crisis: data + PDE + BC weights
- NTK-balanced weighting
- Gradient pathologies and adaptive weights
- Causality weighting for time-domain PINNs
- Curriculum and multi-stage training
- Adaptive collocation point sampling (RAR/RAD)
- Domain decomposition: XPINN, cPINN, FBPINN
Part 5: Part 4: Wave equations in PINN form
Part 6: Part 5: Forward modelling and where PINNs fall short
Part 7: Part 6: Velocity inversion with PINNs
- The classical FWI loss vs the PINN-FWI loss
- 1D velocity inversion end-to-end
- 2D Marmousi-class inversion: setup
- Multi-scale frequency continuation
- Cycle skipping: detection and remedies
- Multi-parameter inversion (vp, vs, rho)
- Source-encoded FWI-PINN
- Loss-weight sensitivity in FWI-PINN
- Defending an inversion run: convergence diagnostics
Part 8: Part 7: Travel-time, surface-wave, and joint inversion
Part 9: Part 8: Operator learning for seismology
Part 10: Part 9: Hybrid PINN + classical, with uncertainty
Part 11: Part 10: Field and benchmark capstones
- Marmousi velocity inversion (the field-standard testbed)
- Microseismic monitoring of a Marcellus frac stage
- USArray surface-wave tomography of the western US
- AVO from PINN-augmented FWI
- Sub-salt imaging at a Gulf-of-Mexico class target
- 4D time-lapse monitoring at a Sleipner CO2-injection target
- Eikonal basin tomography
- Reproducing a current arXiv PINN paper end-to-end
Part 12: Part 11: Self-assessment quizzes
- Quiz: Part 0: Neural networks from zero
- Quiz: Part 1: PINN formulation
- Quiz: Part 2: Architectures
- Quiz: Part 3: Training pathologies
- Quiz: Part 4: Wave equations
- Quiz: Part 5: Forward modelling
- Quiz: Part 6: Velocity inversion
- Quiz: Part 7: Travel-time and joint inversion
- Quiz: Part 8: Operator learning
- Quiz: Part 9: Hybrid PINN and uncertainty
- Quiz: Part 10: Capstones
- Final exam, integrated assessment