Creates matrix factorization model which could be solved with Alternating Least Squares (Weighted ALS for implicit feedback). For implicit feedback see "Collaborative Filtering for Implicit Feedback Datasets" (Hu, Koren, Volinsky). For explicit feedback model is classic model for rating matrix decomposition with MSE error (without biases at the moment). These two algorithms are proven to work well in recommender systems.
new()creates WRMF model
WRMF$new(
rank = 10L,
lambda = 0,
init = NULL,
preprocess = identity,
feedback = c("implicit", "explicit"),
non_negative = FALSE,
solver = c("conjugate_gradient", "cholesky"),
cg_steps = 3L,
precision = c("double", "float"),
...
)ranksize of the latent dimension
lambdaregularization parameter
initinitialization of item embeddings
preprocessidentity() by default. User spectified function which will
be applied to user-item interaction matrix before running matrix factorization
(also applied during inference time before making predictions).
For example we may want to normalize each row of user-item matrix to have 1 norm.
Or apply log1p() to discount large counts.
This corresponds to the "confidence" function from
"Collaborative Filtering for Implicit Feedback Datasets" paper.
feedbackcharacter - feedback type - one of c("implicit", "explicit")
non_negativelogical, whether to perform non-negative factorization
solvercharacter - solver for "implicit feedback" problem.
One of c("conjugate_gradient", "cholesky").
Usually approximate "conjugate_gradient" is significantly faster and solution is
on par with "cholesky"
cg_stepsinteger > 0 - max number of internal steps in conjugate gradient
(if "conjugate_gradient" solver used). cg_steps = 3 by default.
Controls precision of linear equation solution at the each ALS step. Usually no need to tune this parameter
precisionone of c("double", "float"). Should embeeding matrices be
numeric or float (from float package). The latter is usually 2x faster and
consumes less RAM. BUT float matrices are not "base" objects. Use carefully.
...not used at the moment
fit_transform()fits the model
WRMF$fit_transform(x, n_iter = 10L, convergence_tol = 0.005, ...)
xinput matrix (preferably matrix in CSC format -`CsparseMatrix`
n_itermax number of ALS iterations
convergence_tolconvergence tolerance checked between iterations
...not used at the moment
transform()create user embeddings for new input
WRMF$transform(x, ...)
xuser-item iteraction matrix
...not used at the moment
clone()The objects of this class are cloneable with this method.
WRMF$clone(deep = FALSE)
deepWhether to make a deep clone.
# NOT RUN {
data('movielens100k')
train = movielens100k[1:900, ]
cv = movielens100k[901:nrow(movielens100k), ]
model = WRMF$new(rank = 5, lambda = 0, feedback = 'implicit')
user_emb = model$fit_transform(train, n_iter = 5, convergence_tol = -1)
item_emb = model$components
preds = model$predict(cv, k = 10, not_recommend = cv)
# }
Run the code above in your browser using DataLab