MLSEC 3

Question 1

Numerical Features

Answer 1

Mapping of events to a vector space

Question 2

Normalization of numerical features

Answer 2

Scaling and offset often different

Standard normalization

Min-max normalization

Question 3

Problems with Numerical Features

Answer 3

Definition of good feature set is hard

Not suitable for structured data

Question 4

Kernel Function

Answer 4

Similarity measure for objects in a domain X

Question 5

Kernel Function Requirement

Answer 5

symmetric

PSD

Question 6

Bag-of-Words Features

Answer 6

Characterization using non-overlapping substrings

Suitable for analysis of strings with known structure

Question 7

N-gram Features

Answer 7

Characterization using “snippets” of length n (n-grams)

Suitable for analysis of strings with unknown structure

Question 8

All-Substring Kernel

Answer 8

Characterization using all possible substrings

More an academic exercise rather than useful

Question 9

Implementation of string features

Answer 9

Explicit but sparse representation of feature vectors

Question 10

Implementation of string kernel

Answer 10

Implicit definition of feature space

Question 11

Suffix Tree

Answer 11

pecahin per 1,2,3,… huruf dari belakang

mulai root terus ikutin dari depan hurufnya