Linear Classification Linear separability Inseparability Real - - PowerPoint PPT Presentation

linear classification linear separability inseparability
SMART_READER_LITE
LIVE PREVIEW

Linear Classification Linear separability Inseparability Real - - PowerPoint PPT Presentation

Linear Classification Linear separability Inseparability Real world problems: there may not exist a hyperplane that separates cleanly Solution to this inseparability problem: map data to higher dimensional space Called the


slide-1
SLIDE 1

Linear Classification

slide-2
SLIDE 2

Linear separability

slide-3
SLIDE 3
  • Real world problems: there may not exist a hyperplane

that separates cleanly

  • Solution to this “inseparability” problem: map data to

higher dimensional space

  • Called the “feature space”, as opposed to the original “input

space”

  • Inseparable training set can be made separable with proper

choice of feature space

Inseparability

slide-4
SLIDE 4

Feature map

slide-5
SLIDE 5

Linear classifier

slide-6
SLIDE 6

Linear classifier

slide-7
SLIDE 7

Good and bad linear classifiers

slide-8
SLIDE 8

Support Vector Machine

Two popular implementations

slide-9
SLIDE 9

Margin

slide-10
SLIDE 10

Margin

slide-11
SLIDE 11

Linear Support Vector Machine

slide-12
SLIDE 12

Inseparable case

slide-13
SLIDE 13

Linear SVM

slide-14
SLIDE 14

Classification vs Regression

Discrete Continuous

slide-15
SLIDE 15

f(x1,x2) =

  • r

y = f(x) = 10.5 x1 x2 x y

slide-16
SLIDE 16

Protein structure prediction as regression

3D coordinates and angles

slide-17
SLIDE 17
slide-18
SLIDE 18

Linear regression

slide-19
SLIDE 19

Nonlinear regression

slide-20
SLIDE 20

When does linear regression work?

slide-21
SLIDE 21

K nearest neighbor regression

slide-22
SLIDE 22

Nearest neighbor regression

slide-23
SLIDE 23

Nearest neighbor regression

slide-24
SLIDE 24

Filling patches in images

slide-25
SLIDE 25

Linear regression

slide-26
SLIDE 26

Linear Regression

slide-27
SLIDE 27

Linear regression

Squared prediction error