CME 250: Homework 1 Part 1

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

Email *

Stanford SUNetID (i.e. the part before your @stanford.edu email, not a number) *

Classification and Regression

Which of the following statements about classification is false?

1 point

We can almost always perform regression by reformulating the problem as a classification task.

K-nearest neighbors and logistic regression are classification methods.

Many methods, including CART (decisions trees) can be used for regression or for classification.

Classification should be used if the variable we wish to predict is categorical or unordered.

Clear selection

Select all of the following that are regression problems.

1 point

A dataset contains information on the daily stock movements of 500 companies over the past 10 years. We want to predict whether each stock will go up or down tomorrow.

A hospital has collected a dataset of X-rays along with radiologists' view of whether each image contains a tumor. We wish to assign each image with a binary label (0 or 1) of whether a tumor exists in each image.

A real estate company has data on the selling price, square footage, number of bedrooms, etc. of residential homes throughout the US, and wants to predict the selling price of a property from the other variables.

The Rotten Tomatoes database contains movie reviews, each designated "Fresh" or "Rotten". We want to build a model that can detect if a movie review is "Fresh" or "Rotten" given the text.

The city’s transportation department is interested in studying the relationship between the temperature and the number of passengers that ride the main bus line.

Model Flexibility / Bias and Variance

For each of the following problem settings, indicate whether we would generally expect the performance of a flexible statistical learning method to be better or worse than an inflexible method. (ISL 2.4.1)

The sample size n is extremely large, and the number of predictors p is small.

1 point

Flexible method will generally perform better

Inflexible method will generally perform better

Clear selection

The number of predictors p is extremely large, and the number of observations n is small.

1 point

Flexible method will generally perform better

Inflexible method will generally perform better

Clear selection

The relationship between the predictors and response is highly non-linear.

1 point

Flexible method will generally perform better

Inflexible method will generally perform better

Clear selection

The variance of the error terms, i.e. epsilon from lecture notes, is extremely high.

1 point

Flexible method will generally perform better

Inflexible method will generally perform better

Clear selection

K-Nearest Neighbors (KNN)

As you decrease the value of the k parameter in your k-nearest neighbors model, your training error continues to improve, but your test error begins to get worse. This is because:

1 point

Your model has too much bias.

Your model has too much variance.

You are underfitting.

You have an insufficiently flexible model.

Clear selection

Let { (1,5), (1,2), (3,4) } be three training observations belonging to class A, and let { (3,1), (5,3) } be two training observations belonging to class B. Predict the class of new observation X = (3,3) using the k-nearest neighbor classifier with k = 3 (and Euclidean distance).

1 point

X belongs to class A.

X belongs to class B.

There is insufficient information to assign X to a class.

Clear selection

Linear Regression

A regression between foot length (response variable in cm) and height (explanatory variable in inches) for a population resulted in the regression equation: y = 10.9 + 0.23 x. What is the residual for a person who is 73 inches tall with a foot length of 29 cm?

1 point

-1.31 cm

0.23 cm

1.31 cm

10.9 cm

27.69 cm

Clear selection

Which of the following is NOT an assumption that justifies the use of linear regression models for prediction or inference?

1 point

The relationship between dependent variables (response) and independent variables (predictors) is linear.

The variance of errors is constant across values of independent variables and predictions.

The effect of a predictor on the response is independent of other predictors.

The predictors and response have a causal relationship.

Clear selection

Course Feedback

How is the pacing of lecture so far?

Too fast

Too slow

Fine with me

Other:

Clear selection

Submit

Clear form

Never submit passwords through Google Forms.

This form was created inside of Stanford University. Report Abuse

Forms