Week 1

Question 1

Name the 3 learners inputs

Answer 1

Domain Set, Label Set, Training Data

Question 2

Define Domain set

Answer 2

An arbitrary set of objects we wish to label

Question 3

Define Label Set

Answer 3

Assign number to label

Question 4

Define Training Data

Answer 4

Input that the learner has access to

Question 5

Measure of success

Answer 5

error of a classifier, the probability that it does not predict the correct label

Question 6

Define overfitting

Answer 6

When hypothesis fits the training data too well

Question 7

Inductive bias

Answer 7

bias toward a particular set of predictors

Question 8

Confidence parameter

Answer 8

probability of getting nonrepresentative sample

Question 9

Accuracy parameter

Answer 9

quality of prediction

Question 10

ERM

Answer 10

Empirical Risk Minimuzation

Question 11

PAC

Answer 11

Probability Approximately Correct

Question 12

K-NN

Answer 12

K Nearest Neighbor

Question 13

k-NN rule summary

Answer 13

assumption that things that look alike must be alike

Question 14

Why do we need parameters for linkage based clustering algorithms

Answer 14

if kept going, it would eventually result in trivial clustering in one large cluster

Question 15

Name the two parameters of clustering needed

Answer 15

Measure or define the disance between clusters, and determine when to stop merging

Question 16

Three types of linkage clustering

Answer 16

Single Linkage, Average Linkage, Max linkage

Question 17

k-means objective function

Answer 17

uses centroid disctance