4. Support Vector Machines. Recall the (soft-margin) SVM primal problem: minimize :l |1n| |2+ 2 6, subject to ,;

4. Support Vector Machines. Recall the (soft-margin) SVM primal problem: ,...,n for i 1 minimize subject to and its dual problem: 1 c illw112+ n 0 for i = 1 (1 Yz [wTxz + b]) 0 for i = 1 maximize subject to (a) (1 point) The primal objective is A. convex B. concave C. neither convex nor concave (b) (2 points) The dual objective is A. convex B. concave C. neither convex nor concave (c) (3 points) c is an important hyperparameter to tune when solving SVMs. If you find that the training error is too large, to reduce training error, you should A. increase c B. decrease c C. do something else because c is irrelevant