Table 4 Parameters used to limit the number of features
From: Investigating ADR mechanisms with Explainable AI: a feasibility study with knowledge graph mining
Parameter | Domain | Description |
|---|---|---|
k | \({\mathbb {N}}^+\) | Maximum length of paths and path patterns |
t | \({\mathbb {N}}\cup \left\{ -1\right\}\) | Maximum depth of generalization |
deg | \({\mathbb {N}}\cup \left\{ -1\right\}\) | Maximum node degree to allow expansion |
\(s_{\text {min}}\) | \({\mathbb {N}}\) | Minimum support for features |
\(s_{\text {max}}\) | \({\mathbb {N}}\) | Maximum support for features |
Undirected | \({\mathbb {B}}\) | Consider the graph undirected or directed |
\(b_{\text {predicates}}\) | List of URIs | Blacklist of predicates not to traverse |
\(b_{\text {exp-types}}\) | List of URIs | Blacklist of types of entities not to traverse |
\(b_{\text {gen-types}}\) | List of URIs | Blacklist of types not to traverse in generalization |
m | \(\{\texttt {{no- }check} , {\texttt {p}}, {\texttt {g}}, {\texttt {m}}, {\texttt {pg}}, \texttt {pgm}\}\) | Domain-dependent filter for features with at least a pathway (p), a gene (g) or a MeSH class (m) |