multiviews: Mult-iview Simulation Data Sets

In SeojinBang/MKKC: Multiple Kernel K-means Clustering

Description

multiviews contains multi-view simulation data used in Bang et. al (2018). The simulation data sets aim to describe how mkkc concatenates multiple views to extract complementary cluster information from the views with noise or redundant information.

Each multi-view simulation data is generated by taking different combinations of four types of views: complete view, partial view 1, partial view 2, and noisy view. The complete view has complete information to distinguish the three clusters from each other. Each the partial view 1 and partial view 2 only conveys partial information so that each view alone cannot completely detect the three clusters. Partial view 1 can detect the first cluster but cannot recognize difference between the second and third clusters while Partial view 2 can detect the third cluster but cannot recognize difference between the first and second cluster. The noisy view is simply composed of noise variables which do not have any information about cluster.

simAnoise, simAredun1, ..., simAredun5 is composed of a complete view and a partial view 1. simAnoise has additional 10 noise variables to the complete view. simAredun1, ..., simAredun5 have additional 5 redundant pairs to the complete view. simAredun1 has the lowest level of redundancy and simAredun5 has the highest level of redundancy.

simBnoise, simBredun1, ..., simBredun5 is composed of partial view 1 and partial view 2. simBnoise has additional 10 noise variables to the partial view 1. simBredun1, ..., simBredun5 have additional 5 redundant pairs to the partial view 1. simBredun1 has the lowest level of redundancy and simBredun5 has the highest level of redundancy.

simCnoise, simCredun1, ..., simCredun5 is composed of partial view 1, partial view 2, and noisy view. simCnoise has additional 10 noise variables to the partial view 1. simCredun1, ..., simCredun5 have additional 5 redundant pairs to the partial view 1. simCredun1 has the lowest level of redundancy and simCredun5 has the highest level of redundancy.

For details, see Bang et. al (2018).

Usage

simAnoise
simBnoise
simCnoise
simAredun1
simAredun2
simAredun3
simAredun4
simAredun5
simBredun1
simBredun2
simBredun3
simBredun4
simBredun5
simCredun1
simCredun2
simCredun3
simCredun4
simCredun5

Format

simAnoise is a list of three elements named view1, view2, and true.label. view1 is a dataframe with 300 cases (rows) and 12 variables (columns) named var1, var2, noise1, ..., noise10. view2 is a dataframe with 300 cases (rows) and 2 variables (columns) named var1 and var2. true.label is a factor-type vector with three levels (100 cases for each).

simBnoise is a list with the same format as simAnoise.

simCnoise is a list of four elements named view1, view2, view3, and true.label. view1, view2, and true.label have the same format as simAnoise. view3 is a dataframe with 300 cases (rows) and 5 variables (columns) named noise1, ..., noise5.

simAredun1, ..., simAredun5 are lists of three elements named view1, view2, and true.label. view1 is a dataframe with 300 cases (rows) and 12 variables (columns) named var1, var2, redun11, redun12 ..., redun51, redun52. view2 is a dataframe with 300 cases (rows) and 2 variables (columns) named var1 and var2. true.label is a factor-type vector with three levels (100 cases for each).

simBredun1, ..., simBredun5 are lists with the same format as simAredun1, ..., simAredun5.

simCredun1, ..., simCredun5 are lists of four elements named view1, view2, view3, and true.label. view1, view2, and true.label have the same format as simAredun1, ... simAredun5. view3 is a dataframe with 300 cases (rows) and 5 variables (columns) named redun11, redun12 ..., redun51, redun52.

References

\insertRef

bang2018mkkcMKKC

Examples

## visualize multi-view simulation data set A having three noise variables
n.noise <- 3      # number of noises to be added
heatmap(simAnoise$view1[,c(1:(2 + n.noise))],
        scale = "column", Rowv = NA, Colv = NA)     # view 1
heatmap(simAnoise$view2,
        scale = "column", Rowv = NA, Colv = NA)     # view 2

## visualize multi-view simulation data set B having three redundant pairs
n.redunpair <- 3  # number of redundant pairs to be added
heatmap(simBredun2$view1[,c(1:(2 + 2 * n.redunpair))],
        scale = "column", Rowv = NA, Colv = NA)      # view 1
heatmap(simBredun2$view2,
        scale = "column", Rowv = NA, Colv = NA)      # view 2

SeojinBang/MKKC documentation built on Sept. 18, 2019, 1:42 p.m.