How qiunao-nzbgy/becat works
BECAT is a plug-and-play graph-domain-adaptation adapter, released alongside nine citation-experiment backbone integrations. The documented workflow is to train a selected backbone on source and target graph data, refine its latent representations with BECAT, and pass refined embeddings to the downstream classifier while retaining the backbone’s original adaptation branch. The shared adapter’s geometry stages are documented in the README; integration call details are not uniformly shown in the sampled excerpts, so those links are marked unverified.
Main entry points: becat.py PyTorch adapter · DANN DANN integration · ASN ASN integration · UDAGCN UDAGCN integration
Components 17
Shared adapters
- Latent graph
becat/torch/becat.py - Curvature signatures
becat/torch/becat.py - Geometry alignment
becat/torch/becat.py - Sparse refinement
becat/torch/becat.py - PyTorch adapter
becat/torch/becat.py - TensorFlow adapter
becat/tensorflow/becat.py
PyTorch integrations
- DANN integration
backbones/DANN - ASN integration
backbones/ASN - UDAGCN integration
backbones/UDAGCN - GRADE integration
backbones/GRADE - SpecReg integration
backbones/SpecReg - PairAlign integration
backbones/PairAlign - A2GNN integration
backbones/A2GNN - DGSDA integration
backbones/DGSDA
TensorFlow integration
- AdaGCN integration
backbones/AdaGCN
Other components
- Experiment runner
- Citation graphs
Connections 23
- Experiment runner → DANN integration: runs transfer
README.md - Experiment runner → ASN integration: runs transfer
README.md - Experiment runner → UDAGCN integration: runs transfer
README.md - Experiment runner → GRADE integration: runs transfer
README.md - Experiment runner → SpecReg integration: runs transfer
README.md - Experiment runner → PairAlign integration: runs transfer
README.md - Experiment runner → A2GNN integration: runs transfer
README.md - Experiment runner → DGSDA integration: runs transfer
README.md - Experiment runner → AdaGCN integration: runs transfer
README.md - Citation graphs → DANN integration: supplies data
README.md - DANN integration → Latent graph: builds from embeddings
- Latent graph → Curvature signatures: supports curvature
README.md - Curvature signatures → Geometry alignment: aligns statistics
README.md - Geometry alignment → Sparse refinement: refines embeddings
README.md - Sparse refinement → DANN integration: feeds classifier
README.md - ASN integration → PyTorch adapter: uses adapter
- UDAGCN integration → PyTorch adapter: uses adapter
- GRADE integration → PyTorch adapter: uses adapter
- SpecReg integration → PyTorch adapter: uses adapter
- PairAlign integration → PyTorch adapter: uses adapter
- A2GNN integration → PyTorch adapter: uses adapter
- DGSDA integration → PyTorch adapter: uses adapter
- AdaGCN integration → TensorFlow adapter: uses adapter