sense making in linear algebra lee peng yee bangkok 10-07-2008

31
Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Upload: adam-bailey

Post on 17-Jan-2018

221 views

Category:

Documents


0 download

DESCRIPTION

Contents Vector spaces and bases Matrices Eigenvalues and eigenvectors

TRANSCRIPT

Page 1: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Sense making in linear algebra

Lee Peng YeeBangkok10-07-2008

Page 2: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Historical events

• Geometry went algebraic after Felix Klein• Algebra turned abstract• Linear algebra came from geometry

Page 3: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Contents

• Vector spaces and bases• Matrices• Eigenvalues and eigenvectors

Page 4: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Two questions

• Why linear algebra or motivation• Why eigenvalues and eigenvectors

Page 5: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Why linear algebra

• Linear systems• Geometric transformations• Markov chains• Lately linear codes

Page 6: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

LINEAR CODES

• Message encode transmit received decode detect error correct error final message

• Concepts used: vector space/linear space, basis, matrices, matrix multiplication

Page 7: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

An example

• {000 000, 001 110, 010 101, 011 011,100 011, 101 101, 110 110, 111 000}a linear code or a linear space (closed under linear combination with 1 + 1 = 0)

• Elements in the space are codewords

Page 8: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Basis

• { 100 011, 010 101, 001 110 } forms a basis for the space{000 000, 001 110, 010 101, 011 011,100 011, 101 101, 110 110, 111 000}

• Check 000 000 = 100 011 + 100 011, 011 011 = 010 101 + 001 110 etc

Page 9: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Generator matrix

011101110

100010001

G

Page 10: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Message

• [110] is a 3-bit message• We turn it into a codeword (encode)• Transmit the codeword• Then decode

Page 11: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Encoding

011011011101110

100010001

011

Page 12: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Message received

• The codeword [110 110] is transmitted • Suppose the received word is

[100 110] (with an error)• [100 110] is not a codeword• How do we decode, detect the error and

correct it?

Page 13: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Parity check matrix

100010001

011101110

H

Page 14: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Decoding

000

011011

100010001

011101110

101

011001

100010001

011101110

Page 15: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Error detecting

• If HxT = [0 0 0]T then x is a codeword • If HxT = [1 0 1]T then en error is detected

• If x is a codeword, r is received word, and e is error then HrT = HxT + HeT = HeT

Page 16: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Error correcting

• [1 0 1]T is the syndrome of the errors• [1 0 0 1 1 0] has an error in the second

entry• The corrected message is [1 1 0 1 1 0]

Page 17: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Summary

• Codewords of length 6• 3-bit messages• At most one error• Use generator matrix to encode and parity

check matrix to decode

Page 18: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Hamming code (1950)

• {0000000, 1101001, 0101010, 1000011, 1001100, 0100101, 1100110, 0001111, 1110000, 0011001, 1011010, 0110011, 0111100, 1010101, 0010110, 1111111} the (7,4) Hamming code

Page 19: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

An application of linear codes

• In 1971 Mariner 9 transmitted pictures of Mars back to earth

• The distance between Mars and earth is 84 million miles

• The transmitter on Mariner 9 had only 20 watts

Page 20: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Why eigenvectors

• Diagonalization• Write A = PDP-1 where D is a diagonal

matrix Then An = PDnP-1 (used in Markov chains)

• Alternatively use geometry

Page 21: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

EIGENVECTORS

11

411

3113

11

211

3113

4 and 2 are called eigenvalues

11

11and are called eigenvectors

Page 22: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

What are they for?

1

33113

11

3113

211

3113

08

11

2211

4

Suppose

1

12

11

13 Then

We reduce matrix multiplication to scalar multiplication.

Page 23: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Geometric meaning

13

01

3113

11

411

3113

Page 24: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Finding eigenvectors using geometry

Page 25: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Finding eigenvectorsusing geometry

13

01

3113

11

411

3113

31

10

3113

22

11

3113

13

01

3113

44

11

3113

31

10

3113

2

21

13113

Page 26: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Using eigenvectors as coordinates

3113

13

08

into maps

3113

21

44

)2(214

maps into

Page 27: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Using eigenvectors as coordinates

Page 28: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Eigenvectors as geometry

• To find eigenvectors is to find new coordinates

• To find new coordinates is to simplify computation

• Linear algebra is by no means abstract

Page 29: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

Two recent reports

• www.ed.gov/MathPanel• www.reform.co.uk/documents/The%20val

ue%20of%20mathematics.pdf

Page 30: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

To teach mathematics is to teach skills and rigour

Page 31: Sense making in linear algebra Lee Peng Yee Bangkok 10-07-2008

END

[email protected]