graph4ai/graph4nlp

Confusion about parameters

smith-co opened this issue · 2 comments

smith-co commented

❓ Questions and Help

I am confused about the following parameters?

    sim_metric_type: 'weighted_cosine'
    num_heads: 1
    top_k_neigh: null
    epsilon_neigh: 0.5
    smoothness_ratio: 0.1
    connectivity_ratio: 0.05
    sparsity_ratio: 0.1
  1. Are these parameters specific to any specific GNN i.e., GCN, GAT, GGNN, or for GraphSage?
  2. Or they are generic and applicable to all?
hugochan commented

❓ Questions and Help

I am confused about the following parameters?

    sim_metric_type: 'weighted_cosine'
    num_heads: 1
    top_k_neigh: null
    epsilon_neigh: 0.5
    smoothness_ratio: 0.1
    connectivity_ratio: 0.05
    sparsity_ratio: 0.1
  1. Are these parameters specific to any specific GNN i.e., GCN, GAT, GGNN, or for GraphSage?
  2. Or they are generic and applicable to all?

@smith-co thanks for your interest to the library! These are parameters related to dynamic graph construction and don't depend on specific GNNs. You might want to refer to library documentation for more details.

AlanSwift commented

This issue will be closed. Feel free to reopen it if needed.