1 of 22

Practice Exam 1, Short Answers

Feb.23.2024

Li Chen, lic2@andrew.cmu.edu

2 of 22

Li Chen, lic2@andrew.cmu.edu

3 of 22

Closed-form of Linear Regression:

Closed-form of Ridge Regression:

Jan.31, Distributed Linear Regression, page 29-32

Li Chen, lic2@andrew.cmu.edu

4 of 22

Closed-form of Ridge Regression:

Jan.31, Distributed Linear Regression, page 36

Li Chen, lic2@andrew.cmu.edu

Time Complexity:

5 of 22

Closed-form of Ridge Regression:

Li Chen, lic2@andrew.cmu.edu

Time Complexity:

6 of 22

Li Chen, lic2@andrew.cmu.edu

7 of 22

Li Chen, lic2@andrew.cmu.edu

8 of 22

Li Chen, lic2@andrew.cmu.edu

9 of 22

Li Chen, lic2@andrew.cmu.edu

Jan.31, Distributed Linear Regression, page 37

10 of 22

Li Chen, lic2@andrew.cmu.edu

11 of 22

Li Chen, lic2@andrew.cmu.edu

12 of 22

Li Chen, lic2@andrew.cmu.edu

13 of 22

(0, 0, 0)

(0, 2, 0)

(3, 2, 0)

(3, 2, 1)

(3, 1, 1)

(3, 2, 1)

Feb.12, Hashing Theory CMS, page 25

Li Chen, lic2@andrew.cmu.edu

14 of 22

Li Chen, lic2@andrew.cmu.edu

15 of 22

HW3

Li Chen, lic2@andrew.cmu.edu

16 of 22

Li Chen, lic2@andrew.cmu.edu

17 of 22

Li Chen, lic2@andrew.cmu.edu

18 of 22

Li Chen, lic2@andrew.cmu.edu

x

Definition of eigenvector & eigenvalue

Multiply A to each side to construct the form AA^T

Plug in L=AA^T; lambda is a scaler

19 of 22

Li Chen, lic2@andrew.cmu.edu

20 of 22

Li Chen, lic2@andrew.cmu.edu

Jan.24, Visualization: PCA, Page 64:

1. We can use random projections by simply multiplying our dataset X by a matrix P where each element of P is chosen randomly from a Gaussian distribution.

​

21 of 22

Li Chen, lic2@andrew.cmu.edu

Jan.24, Visualization: PCA, Page 64:

2.

RP is data independent;

RP supports online-style learning;

RP has better worst-case performance than PCA. (One point can destroy PCA)

​

22 of 22

Li Chen, lic2@andrew.cmu.edu

Jan.29, Visualization: JL + t-SNE