Questions: Kernel Methods and the Kernel Trick

5 questions to test your understanding

Score: 0 / 5
Question 1 Multiple Choice

A machine learning engineer applies an RBF kernel SVM to a dataset that is not linearly separable in 2D. What actually happens during training that allows the SVM to find a non-linear decision boundary?

AThe data points are explicitly transformed to an infinite-dimensional space, and a standard linear SVM is trained there
BThe SVM retrains multiple times with different hyperplane orientations until a curved boundary is discovered
CEvery dot product in the SVM optimization is replaced by a kernel evaluation k(x_i, x_j), so the algorithm behaves as if it is operating in a high-dimensional space while computing only in the original space
DThe RBF kernel compresses the data into a lower-dimensional space where classes become linearly separable
Question 2 Multiple Choice

Why does the polynomial kernel k(x, y) = (x·y + 1)² provide a computational advantage over explicitly mapping data points to the polynomial feature space before computing dot products?

AThe polynomial kernel produces more accurate results than explicit feature mapping because it avoids rounding errors
BThe kernel function computes exactly the same value as the dot product in the expanded feature space, but using only the original coordinates — avoiding the cost of constructing and storing the high-dimensional feature vectors
CThe polynomial kernel reduces the data's dimensionality as a form of implicit regularization
DThe kernel function uses matrix decomposition to skip the dot product computation entirely
Question 3 True / False

The kernel trick can be applied to any machine learning algorithm to make it work in high-dimensional feature spaces.

TTrue
FFalse
Question 4 True / False

Increasing the γ (gamma) parameter of an RBF kernel in an SVM generally produces a smoother, simpler decision boundary.

TTrue
FFalse
Question 5 Short Answer

Explain what it means for a kernel function to 'implicitly compute a dot product in a high-dimensional feature space,' and why this matters for learning non-linear decision boundaries.

Think about your answer, then reveal below.