Objectives

An objective simply consists in a loss function and a decoder. Node-level objectives compute a loss for every node in a time-window graph, whereas edge-level ones compute loss for all edges. This makes node-level objectives usually faster but less powerful than edge-level objectives to capture pair-wise information.

Arguments

• predict_edge_type
  • decoder: str (1)
  • balanced_loss: bool
  • use_triplet_types: bool
• predict_node_type
  • decoder: str (2)
  • balanced_loss: bool
• predict_masked_struct
  • loss: str (3)
  • decoder: str (4)
  • balanced_loss: bool
• detect_edge_few_shot
  • decoder: str (5)
• predict_edge_contrastive
  • decoder: str (6)
• reconstruct_node_features
  • loss: str (7)
  • decoder: str (8)
• reconstruct_node_embeddings
  • loss: str (9)
  • decoder: str (10)
• reconstruct_edge_embeddings
  • loss: str (11)
  • decoder: str (12)
• reconstruct_masked_features
  • loss: str (13)
  • mask_rate: float
  • decoder: str (14)
• one_class
  • decoder: str (15)
  • beta: float (16)
  • eps: float (17)
  • warmup: int (18)

1. Decoder used before computing loss.
   
   Available options (one selection):
   edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
2. Decoder used before computing loss.
   
   Available options (one selection):
   edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
3. 
   Available options (one selection):
   cross_entropy
BCE
4. Decoder used before computing loss.
   
   Available options (one selection):
   edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
5. Decoder used before computing loss.
   
   Available options (one selection):
   edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
6. Decoder used before computing loss.
   
   Available options (one selection):
   edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
7. 
   Available options (one selection):
   SCE
MSE
MSE_sum
MAE
none
8. Decoder used before computing loss.
   
   Available options (one selection):
   edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
9. 
   Available options (one selection):
   SCE
MSE
MSE_sum
MAE
none
10. Decoder used before computing loss.
    
    Available options (one selection):
    edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
11. 
    Available options (one selection):
    SCE
MSE
MSE_sum
MAE
none
12. Decoder used before computing loss.
    
    Available options (one selection):
    edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
13. 
    Available options (one selection):
    SCE
MSE
MSE_sum
MAE
none
14. Decoder used before computing loss.
    
    Available options (one selection):
    edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
15. Decoder applied to embeddings before the hypersphere; use 'none' (identity).
    
    Available options (one selection):
    edge_mlp
node_mlp
magic_gat
nodlink
inner_product
none
16. Soft-boundary fraction; radius is the (1-beta) distance quantile.
    
17. Center slack keeping |c| away from zero.
    
18. Kept for parity; a no-op (c/r update every train step).