Change refit.tidylda defaults

TommyJones / tidylda

Implements an algorithim for Latent Dirichlet Allocation using style conventions from the [tidyverse](https://style.tidyverse.org/) and [tidymodels](https://tidymodels.github.io/model-implementation-principles/index.html).

Other

41 stars 3 forks source link

Change refit.tidylda defaults #49

Closed TommyJones closed 3 years ago

TommyJones commented 3 years ago

New arguments (and new order)

object
new_data
iterations
burnin
prior_weight
additional_k
optimize_alpha
calc_likelihood
calc_r2
return_data
threads
verbose

prior_weight is based on my own calculations where

prior_weight = 1 ==> it's like you're just adding new documents to the model
prior_weight > 1 ==> new data takes precedence in refitting the model
prior_weight < 1 ==> the base model's data has more weight than new data

TommyJones commented 3 years ago

https://github.com/TommyJones/tidylda/commit/067d090bd03358e2cb30f7ae3e9f1a2fb9489dd8