For each pair the edge features along the path are weighted by
w_table[step] and averaged across the path length. Bond
type is a property of neither endpoint, so it has to enter the
model here.
Usage
morie_grphmr_edge(paths, edge_features, w_table)
Arguments
- paths
Named list of integer vectors, one per (i, j) pair.
- edge_features
Named list of numeric vectors keyed by edge
"u,v" (and "v,u").
- w_table
Length-K numeric vector of step weights.
Value
A list with edge_bias (named numeric vector),
note.
Examples
# each path is the list of edges walked from i to j
paths <- list(`1,3` = list(c(1, 2), c(2, 3)))
ef <- list(`1,2` = c(1, 0), `2,3` = c(0, 1))
morie_grphmr_edge(paths, ef, w_table = c(0.5, 0.5))$edge_bias
#> $`1,3`
#> [1] 0.25
#>
