The hLDA R package is a wrapper around the HLDA class/functions of
the tomotopy
python library. It allows to fit hierarchical topic models
(hierarchical Latent Dirichlet Allocation or hLDA) on matrix of count
data where each row is a sample (e.g. a document) and each column is a
feature (e.g. a word). Element (i,j) of the count matrix provides the
number of time a given feature j was found in document i.
You can install the development version of hLDA with:
devtools::install_github("lasy/hLDA")This is a basic example which shows you how to fit a hierarchical LDA model to count data and to visualized the fitted parameters. This example fit hLDA to random data, where no structure is expected.
library(hLDA)
set.seed(1)
M <- 10 # number of samples (e.g. documents)
N <- 20 # number of features (e.g. words)
x <- matrix(sample(0:100, N*M, replace = TRUE), M, N)
colnames(x) <- paste0(sample(letters, N, replace = TRUE), 1:N) # random feature names
m <- hLDA(x, depth = 3) # fitting the hLDA to the dataplot_hLDA(m, title = "test with random data")