Skip to contents

Convert Poisson Matrix Factorization to Multinomial Topic model. The same idea was first coined by fastTopics paper.

Usage

pmf2topic(.beta, .theta, eps = 1e-08)

Arguments

.beta

D x K dictionary matrix

.theta

N x K sample-wise factor loading matrix

Value

a list of (beta, prop, depth), where the beta is a re-scaled dictionary matrix, each row of the prop matrix corresponds to a mixing proportion per column, and the depth parameter gauges a sequencing depth for each column

Details

EY|.beta, .theta = .beta %*% t(.theta)