bpr_log_likelihood {BPRMeth} | R Documentation |
These functions evaluate the model log-likelihood and gradient for different observation models. Available models are "bpr" (i.e. "bernoulli" or "binomial"), "beta" and "lr" (i.e. "gaussian"). There are also functions to compute the sum and weighted sum of the observation model likelihoods, e.g. required for the EM algorithm. These functions are written in C++ for efficiency.
bpr_log_likelihood(w, X, H, lambda, is_nll) bpr_gradient(w, X, H, lambda, is_nll) betareg_log_likelihood(w, X, H, lambda, is_nll) betareg_gradient(w, X, H, lambda, is_nll) sum_weighted_bpr_lik(w, X_list, H_list, r_nk, lambda, is_nll) sum_weighted_bpr_grad(w, X_list, H_list, r_nk, lambda, is_nll) sum_weighted_betareg_lik(w, X_list, H_list, r_nk, lambda, is_nll) sum_weighted_betareg_grad(w, X_list, H_list, r_nk, lambda, is_nll) lr_log_likelihood(w, X, H, lambda = 0.5, is_nll = FALSE)
w |
A vector of parameters (i.e. coefficients of the basis functions) |
X |
An |
H |
The |
lambda |
The complexity penalty coefficient for penalized regression. |
is_nll |
Logical, indicating if the Negative Log Likelihood should be returned. |
X_list |
A list of elements of length N, where each element is an
|
H_list |
A list of elements of length N, where each element contains the
|
r_nk |
A vector of length N containing the posterior probabilities (i.e.
responsibilities) for each element of |
Returns the log likelihood or gradient of the observation model.
C.A.Kapourani C.A.Kapourani@ed.ac.uk
eval_functions
, infer_profiles_mle