ELEC 425 Quiz 1 Review

Which statement(s) about validation and cross-validation is/are true? (Select all correct statements.)

1. In some situation you should use validation but not cross validation, e.g., when training takes too long (e.g., months). 2. A validation set can be used to prevent overfitting.

What model(s) listed below is/are discriminative model(s)? (Select all correct answers.)

1. Perceptron 2. Logistic Regression

Which of the following statements about Perceptron is/are true? (Select all correct answers.)

1. The perceptron algorithm can be kernalized; i.e., it can be rewritten to a form that depends only on dot-product of datapoints. 2. For data that are linearly separable, perceptron always converge, i.e., the training process can find a decision boundary in a finite number of steps.

What statement(s) is/are true about a KNN classifier? (Please select all correct answers.)

1. Training is easy---KNN just remembers the training data 2. KNN is a parametric model 3. During test, KNN may need to calculate the distance between a test data point and every training data point. 4. It may consume lots of memory or disk space to save all training data

Nonparametric models refer to machine learning models that do not have parameters. True or False?

False

Suppose our data points are 2-dimensional and we have two data points x = (x1, x2) and z = (z1, z2) and we have the following kernel function k(x, z)= (2*(x^T)z)^2 If we use the above kernel to replace dot product in a kernelized algorithm, this implicitly maps the data point x from (x1, x2) to a higher dimensional image (2*x1^2, 2*x2^2, √ 2*x1 *x2). True or False?

False

The difference between classification and regression is that classification uses categorical inputs/features but regression uses continuous inputs/features. True or False?

False

When overfitting happens, the training error is high and test error is low. True or False?

False

Which statement about Decision Tree is true:

One major goal of training is to pick the order of features to split training data.

Which statement is true about Naive Bayes classifiers

The assumption used in Naive Bayes can be extended to deal with features that have continuous values; an example is the Gaussian Naive Bayes classifier.

What is the main difference between supervised learning and unsupervised learning?

Unsupervised learning learns from data that has not been labeled, classified, or categorized.

If you have trained a Logistic Regression model and got the following hyperplane: wTx = (0.2, −0.5, 0.05)(1, x1, x2)T During test, this can be used to compute p(C1|x) with a logistic sigmoid function acting on it. If you are given a data point x = (x1 = 0.4, x2 = 3), which of the following prediction is correct. Use sigmoid function

p(C1|x) > 0.5, so predict x to be C1