How chenyaojie1263/hybrid-supervised-inversion-method works
This repository implements a hybrid supervised prestack-inversion method through two training workflows. In the supervised workflow, training batches seismic data, labels, and low-frequency models, then optimizes a U-Net/GRU model against the labels. In the unsupervised workflow, training uses seismic data, prior predictions, low-frequency models, and a forward operator to define its loss. Both scripts save model checkpoints. The README identifies the workflows; the unsupervised training excerpt provides the clearest evidence of its loss and operator use.
Main entry points: Unsupervised learning training.py Unsupervised training · Supervised learning training.py Supervised training · Unet_GRU_MODEL.py Supervised U-Net GRU · Unet_GRU_MODEL_Unsupervised.py Unsupervised U-Net GRU
Components 10
Supervised workflow
- Supervised training
Supervised learning training.py - Supervised U-Net GRU
Unet_GRU_MODEL.py - Supervised checkpoint
Supervised learning training.py
Unsupervised workflow
- Unsupervised training
Unsupervised learning training.py - Unsupervised U-Net GRU
Unet_GRU_MODEL_Unsupervised.py - Unsupervised checkpoint
Unsupervised learning training.py
Data inputs
- Supervised dataset
DATA.py - Unsupervised dataset
DATA_un.py - Forward operator
Other components
- Researcher
Connections 11
- Researcher → Supervised training: starts training
README.md - Researcher → Unsupervised training: starts training
README.md - Supervised training → Supervised dataset: batches data
Supervised learning training.py - Supervised training → Supervised U-Net GRU: trains model
Supervised learning training.py - Supervised U-Net GRU → Supervised training: returns prediction
Supervised learning training.py - Supervised training → Supervised checkpoint: saves weights
Supervised learning training.py - Unsupervised training → Unsupervised dataset: batches data
Unsupervised learning training.py - Unsupervised training → Unsupervised U-Net GRU: trains model
Unsupervised learning training.py - Unsupervised U-Net GRU → Unsupervised training: returns prediction
Unsupervised learning training.py - Unsupervised training → Forward operator: applies operator
Unsupervised learning training.py - Unsupervised training → Unsupervised checkpoint: saves weights
Unsupervised learning training.py