Data Science Supervised Learning Algorithms 4

Question 1

What is the bias-variance tradeoff?

Accepted Answer

Balancing underfitting (high bias) against overfitting (high variance)

Answer

Balancing training speed and accuracy

Answer

Choosing between classification and regression

Answer

Selecting the learning rate

Question 2

Why is cross-validation preferred over a single train-test split?

Accepted Answer

It gives a more robust estimate of model performance across folds

Answer

It uses less data

Answer

It eliminates overfitting entirely

Answer

It is faster to compute

Question 3

In multiclass classification, what does the softmax function produce?

Accepted Answer

A probability distribution over all classes summing to 1

Answer

A single binary output

Answer

The largest feature value

Answer

A regression estimate

Question 4

What is a key disadvantage of k-nearest neighbors at prediction time?

Accepted Answer

It is computationally expensive because it stores and searches all training data

Answer

It cannot handle classification

Answer

It requires gradient descent

Answer

It cannot use distance metrics

Question 5

Which scenario most likely indicates an underfitting model?

Accepted Answer

High training error and high test error

Answer

Low training error and high test error

Answer

Low training error and low test error

Answer

Perfect training accuracy

Question 6

What does the ROC curve plot?

Accepted Answer

True positive rate against false positive rate

Answer

Precision against recall

Answer

Accuracy against threshold

Answer

Loss against epochs

Question 7

In gradient descent, what happens if the learning rate is set too high?

Accepted Answer

The algorithm may overshoot and diverge

Answer

Convergence becomes slow but stable

Answer

It always reaches the global minimum faster

Answer

It has no effect on convergence