Determines the latent factors for new users (rows) given their counts for existing items (columns).
This function will use the same method and hyperparameters with which the model was fit. If using this for recommender systems, it's recommended to use instead the function factors.single, even though the new factors obtained with that function might not end up being in the same scale as the original factors in 'model$A'.
Note that, when using proximal gradient method (the default), results from this function and from ‘get.factor.matrices' on the same data might differ a lot. If this is a problem, it’s recommended to use conjugate gradient instead.
A Poisson factorization model as returned by 'poismf'.
New data for whose rows to determine latent factors. Can be passed as a 'data.frame' or as a sparse or dense matrix (see documentation of poismf for details on the data type). While other functions only accept sparse matrices in COO (triplets) format, this function will also take CSR matrices from the 'SparseM' and 'Matrix' packages (classes 'dgRMatrix'/'RsparseMatrix' for 'Matrix'). Inputs will be converted to CSR regardless of their original format.
Note that converting a matrix to 'dgRMatrix' format might require using 'as(m, "RsparseMatrix")' instead of using 'dgRMatrix' directly.
If passing a 'data.frame', the first column should contain row indices or IDs, and these will be internally remapped - the mapping will be available as the row names for the matrix if passing 'add_names=TRUE', or as part of the outputs if passing 'add_names=FALSE'. The IDs passed in the first column will not be matched to the existing IDs of 'X' passed to 'poismf'.
If 'X' passed to 'poismf' was a 'data.frame', 'X' here must also be passed as 'data.frame'. If 'X' passed to 'poismf' was a matrix and 'X' is a 'data.frame', the second column of 'X' here should contain column numbers (with numeration starting at 1).
Whether to add row names to the output matrix if the indices were internally remapped - they will only be so if the 'X' here is a 'data.frame'. Note that if the indices in passed in 'X' here (first and second columns) are integers, once row names are added, subsetting 'X' by an integer will give the row at that position - that is, if you want to obtain the corresponding row for ID=2 from 'X' in 'A_out', you need to use 'A_out["2", ]', not 'A_out[2, ]'.
The factors are initialized to the mean of each column in the fitted model.
If 'X' was passed as a matrix, will output a matrix of dimensions (n, k) with the obtained factors. If passing 'add_names=TRUE' and 'X' passed to 'poismf' was a 'data.frame', this matrix will have row names. Careful with subsetting with integers (see documentation for 'add_names').
If 'X' was passed as a 'data.frame' and passing 'add_names=FALSE' here, will output a list with an entry 'factors' containing the latent factors as described above, and an entry 'mapping' indicating to which row ID does each row of the output correspond.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.