nll: Risk for Cross-Entropy Loss
In nhejazi/nima: Nima Hejazi's R Toolbox

Compute the empirical risk under cross-entropy loss for binary predictions.

1	nll(prediction, outcome)

`prediction`	A `numeric` vector of predicted probabilities.
`outcome`	A `numeric` vector of binary outcomes actually observed.

n_obs <- 100
x <- rnorm(n_obs)
y <- rbinom(n_obs, 1, plogis(x^2))
test_x <- rnorm(n_obs)
test_y <- rbinom(n_obs, 1, plogis(test_x^2))
mod <- glm(y ~ x, family = "binomial")
pred <- predict(mod, newx = as.data.frame(test_x), type = "response")
error <- nll(prediction = unname(pred), outcome = test_y)