qbiocode.apps.quvine.baselines package#
Optional baseline embedding methods.
Every name below is re-exported for convenience only. The guards keep
import qbiocode working when a baseline’s dependency is absent, but they do
not make the method registry resilient – adapters imports from each
module directly, so a genuinely broken baseline still fails loudly there, which
is the correct outcome. What the guards must not do is fail silently: a
shorter __all__ with no explanation is undiagnosable, so each failure is
recorded in UNAVAILABLE and logged at DEBUG.
Submodules:
- qbiocode.apps.quvine.baselines.adapters module
- qbiocode.apps.quvine.baselines.appnp module
- qbiocode.apps.quvine.baselines.configs module
- qbiocode.apps.quvine.baselines.gat module
- qbiocode.apps.quvine.baselines.gcn_mf module
- qbiocode.apps.quvine.baselines.graphgps module
- qbiocode.apps.quvine.baselines.graphsage module
- qbiocode.apps.quvine.baselines.hyperparameter_loader module
- qbiocode.apps.quvine.baselines.netmf module
- qbiocode.apps.quvine.baselines.node2vec module
- qbiocode.apps.quvine.baselines.registration module
- qbiocode.apps.quvine.baselines.registry module
Summary#
__all__ Classes:
Single Graph Convolutional Layer. |
|
GCN-MF: Graph Convolutional Network with Matrix Factorization. |
|
Hyperparameters for the GraphGPS encoder. |
|
Training config for unsupervised link reconstruction or node classification. |
|
QuVINE GCN-MF: GCN-MF with precomputed quantum-calibrated features. |
__all__ Functions:
Generate node embeddings using APPNP. |
|
Generate baseline graph filter embeddings (WITHOUT quantum calibration). |
|
Generate baseline GCN-MF embeddings (WITHOUT quantum calibration). |
|
Generate embeddings using one feature variant followed by GraphGPS. |
|
Normalize adjacency matrix: D^(-1/2) * A * D^(-1/2) |
|
Precompute quantum-calibrated diffusion OFFLINE (before training). |
|
Run APPNP and return embeddings aligned to nodes. |
|
Unsupervised GraphSAGE embedding. |
|
Run NetMF (Network Embedding as Matrix Factorization) on a graph. |
|
Run Node2Vec and return embeddings aligned to nodes. |
|
Train GCN-MF model. |
__all__ Data:
Reference#
- UNAVAILABLE: dict[str, str] = {}#
Maps a baseline module name to the reason it could not be imported. Empty on a complete install. Inspect it when an expected method is missing from
qbiocode.apps.quvine.list_methods().