4. We estimate the logistic regression coefficient vector by minimizing J() = (-log) likelihood function. Which three of the following statements about this cost function JO) are correct? The MLE of '3, i.e., the minimizer of J(), may not exist. The cost function J() for logistic regression is convex, so any local minimum is a global minimum. The Newton-Raphson algorithm, which we use to find the minimizer of JO), could get stuck at a local minimum, even if the global minimum exists. The cost function J() for logistic regression is always non-negative. The cost function J() for logistic regression is only positive at some values, e.g., at the MLE of 93. 1 point