qbiocode.apps.quvine.baselines.configs module

Configuration dataclasses for all baseline methods.

This module provides type-safe configuration classes for each baseline method, eliminating the need for getattr calls and providing clear defaults.

Summary

Classes:

APPNPConfig	Configuration for APPNP baseline.
BaselineFilterConfig	Configuration for baseline filter methods (heat/poly without quantum calibration).
GATMethodConfig	Complete configuration for GAT-based methods.
GATModelConfig	Model architecture configuration for GAT.
GATTrainConfig	Training configuration for GAT models.
GCNMFConfig	Configuration for GCN-MF baseline.
GraphGPSMethodConfig	Complete configuration for GraphGPS-based methods.
GraphGPSModelConfig	Model architecture configuration for GraphGPS.
GraphGPSTrainConfig	Training configuration for GraphGPS models.
GraphSAGEConfig	Configuration for GraphSAGE baseline.
NetMFConfig	Configuration for NetMF baseline.
Node2VecConfig	Configuration for Node2Vec baseline.
QuvineFilterConfig	Configuration for quantum-calibrated filter methods.
QuvineGCNMFConfig	Configuration for quantum-calibrated GCN-MF methods.
QuvineSGNSConfig	Configuration for the QuVINE SGNS walk embeddings (quvine_rwr/ctqw/dtqw).

Functions:

build_appnp_config	Build APPNP config from OmegaConf.
build_baseline_filter_config	Build baseline filter config from OmegaConf.
build_gat_config	Build GAT method config from OmegaConf.
build_gcnmf_config	Build GCN-MF config from OmegaConf.
build_graphgps_config	Build GraphGPS method config from OmegaConf.
build_graphsage_config	Build GraphSAGE config from OmegaConf.
build_netmf_config	Build NetMF config from OmegaConf.
build_node2vec_config	Build Node2Vec config from OmegaConf.
build_quvine_filter_config	Build quantum-calibrated filter config from OmegaConf.
build_quvine_gcnmf_config	Build quantum-calibrated GCN-MF config from OmegaConf.
build_quvine_sgns_config	Build config for a QuVINE SGNS walk embedding (quvine_rwr/ctqw/dtqw).
get_embedding_dim	Get embedding dimension with fallback logic.

Reference

class Node2VecConfig(enabled=False, dimensions=128, walk_length=80, num_walks=10, p=1.0, q=1.0, window=10, min_count=1, workers=4, seed=None)

Bases: object

Configuration for Node2Vec baseline.

enabled: bool = False

dimensions: int = 128

walk_length: int = 80

num_walks: int = 10

p: float = 1.0

q: float = 1.0

window: int = 10

min_count: int = 1

workers: int = 4

seed: Optional[int] = None

class APPNPConfig(enabled=False, dimensions=128, hidden_dim=64, n_layers=2, alpha=0.1, K=10, dropout=0.5, lr=0.01, weight_decay=0.0005, epochs=200, seed=None)

Bases: object

Configuration for APPNP baseline.

enabled: bool = False

dimensions: int = 128

hidden_dim: int = 64

n_layers: int = 2

alpha: float = 0.1

K: int = 10

dropout: float = 0.5

lr: float = 0.01

weight_decay: float = 0.0005

epochs: int = 200

seed: Optional[int] = None

class BaselineFilterConfig(enabled=False, filter_type='heat', embedding_dim=128, t=1.0, K=4, normalize=True, use_features=False, features=None, random_state=None)

Bases: object

Configuration for baseline filter methods (heat/poly without quantum calibration).

enabled: bool = False

filter_type: str = 'heat'

embedding_dim: int = 128

t: float = 1.0

K: int = 4

normalize: bool = True

use_features: bool = False

features: Optional[Any] = None

random_state: Optional[int] = None

class GCNMFConfig(enabled=False, embedding_dim=128, hidden_dim=64, mf_dim=64, n_layers=2, epochs=200, lr=0.01, weight_decay=0.0005, random_state=None)

Bases: object

Configuration for GCN-MF baseline.

enabled: bool = False

embedding_dim: int = 128

hidden_dim: int = 64

mf_dim: int = 64

n_layers: int = 2

epochs: int = 200

lr: float = 0.01

weight_decay: float = 0.0005

random_state: Optional[int] = None

class GATTrainConfig(epochs=200, lr=0.005, weight_decay=0.0005, patience=25, edge_batch_size=4096, val_edge_fraction=0.1, device='cpu', random_state=None, verbose=False)

Bases: object

Training configuration for GAT models.

epochs: int = 200

lr: float = 0.005

weight_decay: float = 0.0005

patience: int = 25

edge_batch_size: int = 4096

val_edge_fraction: float = 0.1

device: str = 'cpu'

random_state: Optional[int] = None

verbose: bool = False

class GATModelConfig(hidden_dim=64, output_dim=128, num_layers=2, heads=4, dropout=0.2, attention_dropout=0.2, negative_slope=0.2, residual=True)

Bases: object

Model architecture configuration for GAT.

hidden_dim: int = 64

output_dim: int = 128

num_layers: int = 2

heads: int = 4

dropout: float = 0.2

attention_dropout: float = 0.2

negative_slope: float = 0.2

residual: bool = True

class GATMethodConfig(enabled=False, variant='raw', embedding_dim=128, heat_t=1.0, poly_K=4, poly_ridge=1e-05, rwr_alpha=0.15, rwr_steps=50, model=<factory>, train=<factory>)

Bases: object

Complete configuration for GAT-based methods.

enabled: bool = False

variant: str = 'raw'

embedding_dim: int = 128

heat_t: float = 1.0

poly_K: int = 4

poly_ridge: float = 1e-05

rwr_alpha: float = 0.15

rwr_steps: int = 50

model: GATModelConfig

train: GATTrainConfig

class GraphGPSTrainConfig(task='link_reconstruction', epochs=200, lr=0.005, weight_decay=0.0005, patience=30, edge_batch_size=8192, val_edge_fraction=0.1, device='cpu', random_state=None, verbose=False)

Bases: object

Training configuration for GraphGPS models.

task: str = 'link_reconstruction'

epochs: int = 200

lr: float = 0.005

weight_decay: float = 0.0005

patience: int = 30

edge_batch_size: int = 8192

val_edge_fraction: float = 0.1

device: str = 'cpu'

random_state: Optional[int] = None

verbose: bool = False

class GraphGPSModelConfig(hidden_dim=64, output_dim=128, num_layers=2, heads=4, dropout=0.2, attn_dropout=0.2, local_gnn='gcn', attn_type='multihead', use_layer_norm=True, activation='relu', lap_pe_dim=0, standardize_features=True)

Bases: object

Model architecture configuration for GraphGPS.

hidden_dim: int = 64

output_dim: int = 128

num_layers: int = 2

heads: int = 4

dropout: float = 0.2

attn_dropout: float = 0.2

local_gnn: str = 'gcn'

attn_type: str = 'multihead'

use_layer_norm: bool = True

activation: str = 'relu'

lap_pe_dim: int = 0

standardize_features: bool = True

class GraphGPSMethodConfig(enabled=False, variant='raw', embedding_dim=128, heat_t=1.0, poly_K=4, poly_ridge=1e-05, rwr_alpha=0.15, rwr_steps=50, model=<factory>, train=<factory>)

Bases: object

Complete configuration for GraphGPS-based methods.

enabled: bool = False

variant: str = 'raw'

embedding_dim: int = 128

heat_t: float = 1.0

poly_K: int = 4

poly_ridge: float = 1e-05

rwr_alpha: float = 0.15

rwr_steps: int = 50

model: GraphGPSModelConfig

train: GraphGPSTrainConfig

class GraphSAGEConfig(enabled=False, dimensions=128, hidden_dim=256, n_layers=2, epochs=50, lr=0.01, neg_samples=5, seed=None)

Bases: object

Configuration for GraphSAGE baseline.

enabled: bool = False

dimensions: int = 128

hidden_dim: int = 256

n_layers: int = 2

epochs: int = 50

lr: float = 0.01

neg_samples: int = 5

seed: Optional[int] = None

class QuvineFilterConfig(enabled=False, filter_type='heat', embedding_dim=128, t=None, K=4, ridge=1e-06, normalize=True, use_features=False, features=None, random_state=None)

Bases: object

Configuration for quantum-calibrated filter methods.

enabled: bool = False

filter_type: str = 'heat'

embedding_dim: int = 128

t: Optional[float] = None

K: int = 4

ridge: float = 1e-06

normalize: bool = True

use_features: bool = False

features: Optional[Any] = None

random_state: Optional[int] = None

class QuvineGCNMFConfig(enabled=False, diffusion_type='heat', embedding_dim=128, hidden_dim=64, mf_dim=64, n_layers=2, epochs=200, lr=0.01, weight_decay=0.0005, K=4, ridge=1e-06, normalize_laplacian=True, random_state=None)

Bases: object

Configuration for quantum-calibrated GCN-MF methods.

enabled: bool = False

diffusion_type: str = 'heat'

embedding_dim: int = 128

hidden_dim: int = 64

mf_dim: int = 64

n_layers: int = 2

epochs: int = 200

lr: float = 0.01

weight_decay: float = 0.0005

K: int = 4

ridge: float = 1e-06

normalize_laplacian: bool = True

random_state: Optional[int] = None

class QuvineSGNSConfig(enabled=False, walk_kind='rwr', cfg=None, n_jobs=1, chunk_size=30)

Bases: object

Configuration for the QuVINE SGNS walk embeddings (quvine_rwr/ctqw/dtqw).

These methods are the core QuVINE embedding: per-root views -> walks (rwr/ctqw/dtqw) -> corpus -> word2vec (SGNS). They do NOT use quantum calibration targets; the walk itself is the quantum component. The full OmegaConf cfg is carried through because the SGNS core reads walks.*, views.*, train.*, min_count and experiment.base_seed from it.

enabled: bool = False

walk_kind: str = 'rwr'

cfg: Any = None

n_jobs: int = 1

chunk_size: int = 30

class NetMFConfig(enabled=False, dimensions=128, window_size=10, rank=256, negative=1, seed=None)

Bases: object

Configuration for NetMF baseline.

enabled: bool = False

dimensions: int = 128

window_size: int = 10

rank: int = 256

negative: int = 1

seed: Optional[int] = None

get_embedding_dim(cfg, method_cfg, default=128)

Get embedding dimension with fallback logic.

Priority: 1. Method-specific embedding_dim 2. Global train.embedding_dim 3. Default value

Return type:

int

build_node2vec_config(cfg, base_seed)

Build Node2Vec config from OmegaConf.

Return type:

Node2VecConfig

build_appnp_config(cfg, base_seed)

Build APPNP config from OmegaConf.

Return type:

APPNPConfig

build_baseline_filter_config(cfg, base_seed, filter_type='heat')

Build baseline filter config from OmegaConf.

Return type:

BaselineFilterConfig

build_gcnmf_config(cfg, base_seed, config_name='baseline_gcnmf')

Build GCN-MF config from OmegaConf.

Return type:

GCNMFConfig

build_gat_config(cfg, base_seed, config_name)

Build GAT method config from OmegaConf.

Return type:

GATMethodConfig

build_graphgps_config(cfg, base_seed, config_name)

Build GraphGPS method config from OmegaConf.

Return type:

GraphGPSMethodConfig

build_graphsage_config(cfg, base_seed)

Build GraphSAGE config from OmegaConf.

Return type:

GraphSAGEConfig

build_quvine_filter_config(cfg, base_seed, config_name, filter_type)

Build quantum-calibrated filter config from OmegaConf.

Return type:

QuvineFilterConfig

build_quvine_gcnmf_config(cfg, base_seed, config_name, diffusion_type)

Build quantum-calibrated GCN-MF config from OmegaConf.

Return type:

QuvineGCNMFConfig

build_quvine_sgns_config(cfg, base_seed, config_name, walk_kind)

Build config for a QuVINE SGNS walk embedding (quvine_rwr/ctqw/dtqw).

The executor runs the shared SGNS core (views -> walks -> corpus -> word2vec) for the single walk_kind, so the whole cfg is carried through. Enabled follows the same cfg.baselines.<name>.enabled pattern as the other methods.

Return type:

QuvineSGNSConfig

build_netmf_config(cfg, base_seed)

Build NetMF config from OmegaConf.

Return type:

NetMFConfig