aibd {aibd}R Documentation

Define an Attraction Indian Buffet Distribution (AIBD) for Feature Allocations

Description

This function specifies an Attraction Indian Buffet Distribution (AIBD), which is a distribution over feature allocations.

Usage

aibd(
  mass,
  permutation,
  temperature,
  distance,
  decayFunction = c("exponential", "reciprocal", "identity")[1]
)

Arguments

mass

The mass (a.k.a., concentration) parameter of the AIBD.

permutation

A permutation, i.e., a vector of integers 1, 2, ..., n whose length is n and whose elements are unique. Using the Indian buffet analogy, the permutation represents the order the customers enter the buffet.

temperature

A nonnegative scalar which determines how influential the distance matrix is in the feature allocation distribution. The AIBD reduces to the IBP when the temperature is zero and diverges from the IBP as the temperature increases.

distance

A distance matrix, i.e., a symmetric matrix whose (i,j) entry is small if items i and j are similar. An object of class "dist" is also permissible.

decayFunction

One of the following strings: "exponential" (making similarity = exp(-temperature*distance)), "reciprocal" (making similarity = 1/distance^temperature), or "identity" (in which case distance is interpreted as a similarity instead of a distance).

Value

An object representing an Attraction Indian Buffet Distribution (AIBD) for feature allocations.

Examples

states <- c("California","Wisconsin","Nebraska","New York")
data <- USArrests[states,]
dist <- dist(scale(data))
aibd(1, seq_along(states), 1.0, dist)


[Package aibd version 0.1.9 Index]