Which one is the correct generalisation $($i$.$e$.,$ without ambiguous regions$)$ of binary$-$class to multi$-$class $($i$.$e$.,$ K classes$)$ problems?

a$.$Use K discriminant functions to represent the K classes. For each data point, choose the class with the largest magnitude of discriminant function.

b$.$Use K$-1$ discriminant classifiers, each separating one class from the rest

c$.$Use K$($K$-1)/2$ discriminant functions, each for every possible pair of classes $($One$-$versus$-$one classifier$)$