What does the kernel trick in SVM facilitate?

Elimination of the need for optimization algorithms

Reduction of computational complexity to linear time

Direct computation of distances in the original feature space

Efficient computation in a high$-$dimensional space without explicitly mapping input features