Podcast
Questions and Answers
A kernel is a dissimilarity metric between vectors.
A kernel is a dissimilarity metric between vectors.
False (B)
The Gram matrix of dot-products between vectors must be negative semidefinite to be valid.
The Gram matrix of dot-products between vectors must be negative semidefinite to be valid.
False (B)
Complex kernels allow us to define simple explicit feature expansions.
Complex kernels allow us to define simple explicit feature expansions.
False (B)
The Gaussian kernel has a finite number of dimensions.
The Gaussian kernel has a finite number of dimensions.
The Taylor series of exp(#) is a polynomial weighted by constants.
The Taylor series of exp(#) is a polynomial weighted by constants.
High degree polynomials are preferred over lower degree polynomials in Gaussian kernels.
High degree polynomials are preferred over lower degree polynomials in Gaussian kernels.
Kernels allow for expressing a wide variety of ideas of similarity in models.
Kernels allow for expressing a wide variety of ideas of similarity in models.
Kernel ridge regression uses a fixed-size kernel across the space.
Kernel ridge regression uses a fixed-size kernel across the space.
In K nearest neighbors regression, only the least similar units are considered.
In K nearest neighbors regression, only the least similar units are considered.
Type-I and Type-II errors are typically considered equally important in decision-making processes.
Type-I and Type-II errors are typically considered equally important in decision-making processes.
We must know the risk score a-priori before estimating it.
We must know the risk score a-priori before estimating it.
Kernel ridge regression is particularly designed to have local variation properties.
Kernel ridge regression is particularly designed to have local variation properties.
