在R中，如何计算两个字符串向量之间的KL距离？

Question

If I have 2 vectors of strings like:

> list1 = c("cat", "dog", "cat", "rabbit", "dog", "cat")
> list2 = c("dog", "rabbit", "dog", "mouse", "dog", "rabbit", "cat")

I can get distributions for each. For example:

> dist1 = table(list1)/length(list1)
> dist2 = table(list2)/length(list2)
> dist1; dist2

list1
      cat       dog    rabbit 
0.5000000 0.3333333 0.1666667 
list2
      cat       dog     mouse    rabbit 
0.1428571 0.4285714 0.1428571 0.2857143 

How do I calculate the KL Distance between these two distributions? (Using dist2 as the baseline.)

The KL functions I've seen (eg, kl.dist) require vectors of the same length.

Answer 1

The following will produce a data frame with one column with the distribution for each vector strings:

library(dplyr)

list1 <- c("cat", "dog", "cat", "rabbit", "dog", "cat")
list2 <- c("dog", "rabbit", "dog", "mouse", "dog", "rabbit", "cat")

dist1 <- table(list1)/length(list1)
dist2 <- table(list2)/length(list2)

BothDist <- full_join(as.data.frame(dist1),as.data.frame(dist2), by = c("list1" = "list2")) 
BothDist[is.na(BothDist)] <- 0

BothDist