laser.embedding.icomp: 300 independent components of LASER representations of tweets...

In haukelicht/politicaltweets: Classify political tweets

300 independent components of LASER representations of tweets in labeled dataset

Description

fastICA object fitted to the LASER embedding representations of tweet texts contained in the training and validation data used to obtain the ensemble classifier distributed with the politicaltweets package

Usage

laser.embedding.icomp

Format

A fastICA object, which is a list with the following elements:

K

The 1024x300 'pre-whitening' matrix projecting embeddings onto the first 300 principal components

W

The 300x300 'un-mixing' matrix estimated from the original model input data

A

The 300x1024 'mixing' matrix estimated from the original model input data

X.means

a double vector with 1024 elements recording embedding vector means in the original model input data

runtime

difftime object reporting the time (in seconds) whitening and un-mixing took

params

a list mapping the parameter used when fitting fastICA

Examples

## Not run: 
 # create fake 'test' observation (1-row matrix)
 X <- matrix(
   rnorm(1024)
   , nrow = 1
   # column names must be "e0001", ..., "e1024"
   , dimnames = list(c("test"), sprintf("e%04d", 1:1024))
 )
 #'  # obtain independent component representation of x
 X %*% laser.embedding.icomp$K %*% laser.embedding.icomp$W

## End(Not run)

haukelicht/politicaltweets documentation built on July 3, 2023, 4:11 a.m.