Linear Models and Intro to Deep Learning

One answer per question, one attempt per quiz — you cannot retake it.

Question 1 of 5

What is the key difference between maximum likelihood (ML) and maximum a posteriori (MAP) estimation?

MAP uses a prior over parameters, while ML does not.MAP is used only for classification tasks, while ML is for regression.MAP requires computing the model evidence, while ML ignores it.MAP always yields more accurate predictions than ML.

Question 2 of 5

Using L2 weight decay (ridge regression) in a linear model corresponds to assuming which prior on the weights?

A Gaussian prior centered at zeroA Laplace prior centered at zeroA uniform prior over a bounded intervalNo prior (pure maximum likelihood)

Question 3 of 5

What loss function does binary logistic regression minimize when using maximum likelihood?

The squared error lossThe cross-entropy lossThe hinge loss functionThe exponential loss function

Question 4 of 5

Why is gradient descent the standard method for training logistic regression models?

There is no closed-form solution for this problemThe closed-form solution is numerically unstableGradient descent guarantees finding the global minimumIt is faster than solving the normal equations directly

Question 5 of 5

Why is a nonlinear activation function essential in a feedforward neural network with multiple layers?

Without nonlinearity, the entire network reduces to a linear modelIt speeds up the convergence of gradient descentIt prevents the model from overfitting the training dataIt encourages the weights to be sparse

0 / 5 answered

Generated from lecture recordings by the autosummary pipeline.