how to multiply · matrix multiplication. given two n-by-n matrices a and b, compute c = ab....

1

How to Multiply

Slides by Kevin Wayne.Copyright © 2005 Pearson-Addison Wesley.All rights reserved.

integers, matrices, and polynomials

2

Complex Multiplication

Complex multiplication. (a + bi) (c + di) = x + yi.

Grade-school. x = ac - bd, y = bc + ad.

Q. Is it possible to do with fewer multiplications?

4 multiplications, 2 additions

3

Complex Multiplication

Complex multiplication. (a + bi) (c + di) = x + yi.

Grade-school. x = ac - bd, y = bc + ad.

Q. Is it possible to do with fewer multiplications?A. Yes. [Gauss] x = ac - bd, y = (a + b) (c + d) - ac - bd.

Remark. Improvement if no hardware multiply.



5.5 Integer Multiplication

5

Addition. Given two n-bit integers a and b, compute a + b.Grade-school. Θ(n) bit operations.

Remark. Grade-school addition algorithm is optimal.

Integer Addition

1

011 1

110 1+

010 1

111

010 1

011 1

100 0

10111

6

Integer Multiplication

Multiplication. Given two n-bit integers a and b, compute a × b.Grade-school. Θ(n2) bit operations.

Q. Is grade-school multiplication algorithm optimal?

1

1

1

0

0

0

1

1

1

0

1

0

1

1

1

0

1

0

1

1

1

1

0

1

00000000

01010101

01010101

01010101

01010101

01010101

00000000

100000000001011

0

1

1

1

1

1

0

0

×

7

To multiply two n-bit integers a and b: Multiply four ½n-bit integers, recursively. Add and shift to obtain result.

Ex.

Divide-and-Conquer Multiplication: Warmup

!

T (n) = 4T n /2( )recursive calls

1 2 4 3 4 + "(n)

add, shift

1 2 3 # T (n) ="(n

2)

!

a = 2n / 2

" a1

+ a0

b = 2n / 2

"b1

+ b0

ab = 2n / 2

" a1+ a

0( ) 2n / 2

"b1

+ b0( ) = 2

n" a

1b

1 + 2

n / 2" a

1b

0+ a

0b

1( ) + a0b

0

a = 10001101 b = 11100001

a1 a0 b1 b0

8

Recursion Tree

!

T (n) =0 if n = 0

4T (n /2) + n otherwise

" # $

n

4(n/2)

16(n/4)

4k (n / 2k)

4 lg n (1)

T(n)

T(n/2)

T(n/4) T(n/4)

T(2) T(2) T(2) T(2) T(2) T(2) T(2) T(2)

T(n / 2k)

T(n/4)

T(n/2)

T(n/4) T(n/4)

T(n/2)

T(n/4) T(n/4)T(n/4)

...

...

!

T (n) = n 2k

k=0

lgn

" = n2

1+ lgn #1

2#1

$

% &

'

( ) = 2n

2 #n

T(n/2)

...

...

......

9

To multiply two n-bit integers a and b: Add two ½n bit integers. Multiply three ½n-bit integers, recursively. Add, subtract, and shift to obtain result.

Karatsuba Multiplication

!

a = 2n / 2

" a1 + a0

b = 2n / 2

"b1 + b0

ab = 2n" a1b1 + 2

n / 2" a1b0 + a0b1( ) + a0b0

= 2n" a1b1 + 2

n / 2" (a1 + a0 ) (b1 +b0 ) # a1b1 # a0b0( ) + a0b0

1 2 1 33

10

To multiply two n-bit integers a and b: Add two ½n bit integers. Multiply three ½n-bit integers, recursively. Add, subtract, and shift to obtain result.

Theorem. [Karatsuba-Ofman 1962] Can multiply two n-bit integersin O(n1.585) bit operations.

Karatsuba Multiplication

!

T (n) " T n /2# $( ) + T n /2% &( ) + T 1+ n /2% &( )recursive calls

1 2 4 4 4 4 4 4 4 3 4 4 4 4 4 4 4 + '(n)

add, subtract, shift

1 2 4 3 4 ( T (n) = O(n

lg 3) = O(n

1.585)

!

a = 2n / 2

" a1 + a0

b = 2n / 2

"b1 + b0

ab = 2n" a1b1 + 2

n / 2" a1b0 + a0b1( ) + a0b0

= 2n" a1b1 + 2

n / 2" (a1 + a0 ) (b1 +b0 ) # a1b1 # a0b0( ) + a0b0

1 2 1 33

11

Karatsuba: Recursion Tree

!

T (n) =0 if n = 0

3T (n /2) + n otherwise

" # $

n

3(n/2)

9(n/4)

T(n)

T(n/2)

T(n/4) T(n/4)T(n/4)

T(n/2)

T(n/4) T(n/4)T(n/4)

T(n/2)

T(n/4) T(n/4)T(n/4)

!

T (n) = n 32( )

k

k=0

lgn

" = n32( )

1+ lgn#1

32#1

$

%

& &

'

(

) ) = 3n

lg 3 #2n

3 lg n (1)T(2) T(2) T(2) T(2) T(2) T(2) T(2) T(2)

T(n / 2k)

...

......

......

3k (n / 2k)

12

Integer division. Given two n-bit (or less) integers s and t,compute quotient q = s / t and remainder r = s mod t.

Fact. Complexity of integer division is same as integer multiplication.To compute quotient q:

Approximate x = 1 / t using Newton's method: After log n iterations, either q = s x or q = s x.

!

xi+1

= 2xi" t x

i

2

Fast Integer Division Too (!)

using fastmultiplication

Matrix Multiplication

14

Dot product. Given two length n vectors a and b, compute c = a ⋅ b.Grade-school. Θ(n) arithmetic operations.

Remark. Grade-school dot product algorithm is optimal.

Dot Product

!

a " b = aibi

i=1

n

#

!

a = .70 .20 .10[ ]

b = .30 .40 .30[ ]a " b = (.70 # .30) + (.20 # .40) + (.10 # .30) = .32

15

Matrix multiplication. Given two n-by-n matrices A and B, compute C = AB.Grade-school. Θ(n3) arithmetic operations.

Q. Is grade-school matrix multiplication algorithm optimal?

Matrix Multiplication

!

cij = aikbkj

k=1

n

"

!

c11

c12

L c1n

c21

c22

L c2n

M M O M

cn1

cn2

L cnn

"

#

$ $ $ $

%

&

' ' ' '

=

a11

a12

L a1n

a21

a22

L a2n

M M O M

an1

an2

L ann

"

#

$ $ $ $

%

&

' ' ' '

(

b11

b12

L b1n

b21

b22

L b2n

M M O M

bn1

bn2

L bnn

"

#

$ $ $ $

%

&

' ' ' '

!

.59 .32 .41

.31 .36 .25

.45 .31 .42

"

#

$ $ $

%

&

' ' '

=

.70 .20 .10

.30 .60 .10

.50 .10 .40

"

#

$ $ $

%

&

' ' '

(

.80 .30 .50

.10 .40 .10

.10 .30 .40

"

#

$ $ $

%

&

' ' '

16

Block Matrix Multiplication

!

C11

= A11"B

11 + A

12"B

21 =

0 1

4 5

#

$ %

&

' ( "

16 17

20 21

#

$ %

&

' ( +

2 3

6 7

#

$ %

&

' ( "

24 25

28 29

#

$ %

&

' ( =

152 158

504 526

#

$ %

&

' (

!

152 158 164 170

504 526 548 570

856 894 932 970

1208 1262 1316 1370

"

#

$ $ $ $

%

&

' ' ' '

=

0 1 2 3

4 5 6 7

8 9 10 11

12 13 14 15

"

#

$ $ $ $

%

&

' ' ' '

(

16 17 18 19

20 21 22 23

24 25 26 27

28 29 30 31

"

#

$ $ $ $

%

&

' ' ' '

C11A11 A12 B11

B11

17

Matrix Multiplication: Warmup

To multiply two n-by-n matrices A and B: Divide: partition A and B into ½n-by-½n blocks. Conquer: multiply 8 pairs of ½n-by-½n matrices, recursively. Combine: add appropriate products using 4 matrix additions.

!

C11

= A11" B

11( ) + A12" B

21( )C

12= A

11" B

12( ) + A12" B

22( )C

21= A

21" B

11( ) + A22" B

21( )C

22= A

21" B

12( ) + A22" B

22( )

!

C11

C12

C21

C22

"

# $

%

& ' =

A11

A12

A21

A22

"

# $

%

& ' (

B11

B12

B21

B22

"

# $

%

& '

!


1 2 4 3 4 + "(n

2)

add, form submatrices

1 2 4 4 3 4 4 # T (n) ="(n

3)

18

Fast Matrix Multiplication

Key idea. multiply 2-by-2 blocks with only 7 multiplications.

7 multiplications. 18 = 8 + 10 additions and subtractions.

!

P1 = A11 " (B12 # B22 )

P2 = (A11 + A12 ) " B22

P3 = (A21 + A22 ) " B11

P4 = A22 " (B21 # B11)

P5 = (A11 + A22 ) " (B11 + B22 )

P6 = (A12 # A22 ) " (B21 + B22 )

P7 = (A11 # A21) " (B11 + B12 )

!

C11

= P5

+ P4" P

2+ P

6

C12

= P1+ P

2

C21

= P3

+ P4

C22

= P5

+ P1" P

3" P

7

!

C11

C12

C21

C22

"

# $

%

& ' =

A11

A12

A21

A22

"

# $

%

& ' (

B11

B12

B21

B22

"

# $

%

& '

19

Fast Matrix Multiplication

To multiply two n-by-n matrices A and B: [Strassen 1969] Divide: partition A and B into ½n-by-½n blocks. Compute: 14 ½n-by-½n matrices via 10 matrix additions. Conquer: multiply 7 pairs of ½n-by-½n matrices, recursively. Combine: 7 products into 4 terms using 8 matrix additions.

Analysis. Assume n is a power of 2. T(n) = # arithmetic operations.

!


1 2 4 3 4 + "(n

2)

add, subtract

1 2 4 3 4 # T (n) ="(n

log2 7) =O(n

2.81)

20

Fast Matrix Multiplication: Practice

Implementation issues. Sparsity. Caching effects. Numerical stability. Odd matrix dimensions. Crossover to classical algorithm around n = 128.

Common misperception. “Strassen is only a theoretical curiosity.” Apple reports 8x speedup on G4 Velocity Engine when n ≈ 2,500. Range of instances where it's useful is a subject of controversy.

Remark. Can "Strassenize" Ax = b, determinant, eigenvalues, SVD, ….

21

Begun, the decimal wars have. [Pan, Bini et al, Schönhage, …]

Fast Matrix Multiplication: Theory

Q. Multiply two 2-by-2 matrices with 7 scalar multiplications?

!

" (nlog3 21) = O(n

2.77)

!

O(n2.7801

)

!

"(nlog2 6) = O(n

2.59)

!

"(nlog2 7 ) =O(n

2.807)A. Yes! [Strassen 1969]

Q. Multiply two 2-by-2 matrices with 6 scalar multiplications?A. Impossible. [Hopcroft and Kerr 1971]

Q. Two 3-by-3 matrices with 21 scalar multiplications?A. Also impossible.

Two 48-by-48 matrices with 47,217 scalar multiplications.

December, 1979.

!

O(n2.521813

)

!

O(n2.521801

) January, 1980.

A year later.

!

O(n2.7799

)

Two 20-by-20 matrices with 4,460 scalar multiplications.

!

O(n2.805

)

22

Fast Matrix Multiplication: Theory

Best known. O(n2.376) [Coppersmith-Winograd, 1987]

Conjecture. O(n2+ε) for any ε > 0.

Caveat. Theoretical improvements to Strassen are progressivelyless practical.

5.6 Convolution and FFT

24

Fourier Analysis

Fourier theorem. [Fourier, Dirichlet, Riemann] Any periodic functioncan be expressed as the sum of a series of sinusoids. sufficiently smooth

t

N = 1N = 5N = 10N = 100

!

y(t) = 2

"

sin kt

kk=1

N

#

25

Euler's Identity

Sinusoids. Sum of sine an cosines.

Sinusoids. Sum of complex exponentials.

eix = cos x + i sin x

Euler's identity

26

Time Domain vs. Frequency Domain

Signal. [touch tone button 1]

Time domain.

Frequency domain.

!

y(t) = 12sin(2" # 697 t) + 1

2sin(2" # 1209 t)

Reference: Cleve Moler, Numerical Computing with MATLAB

frequency (Hz)

amplitude

0.5

time (seconds)

soundpressure

27

Time Domain vs. Frequency Domain

Signal. [recording, 8192 samples per second]

Magnitude of discrete Fourier transform.

Reference: Cleve Moler, Numerical Computing with MATLAB

28

Fast Fourier Transform

FFT. Fast way to convert between time-domain and frequency-domain.

Alternate viewpoint. Fast way to multiply and evaluate polynomials.

If you speed up any nontrivial algorithm by a factor of amillion or so the world will beat a path towards findinguseful applications for it. -Numerical Recipes

we take this approach

29

Fast Fourier Transform: Applications

Applications. Optics, acoustics, quantum physics, telecommunications, radar,

control systems, signal processing, speech recognition, datacompression, image processing, seismology, mass spectrometry…

Digital media. [DVD, JPEG, MP3, H.264] Medical diagnostics. [MRI, CT, PET scans, ultrasound] Numerical solutions to Poisson's equation. Shor's quantum factoring algorithm. …

The FFT is one of the truly great computationaldevelopments of [the 20th] century. It has changed theface of science and engineering so much that it is not anexaggeration to say that life as we know it would be verydifferent without the FFT. -Charles van Loan

30

Fast Fourier Transform: Brief History

Gauss (1805, 1866). Analyzed periodic motion of asteroid Ceres.

Runge-König (1924). Laid theoretical groundwork.

Danielson-Lanczos (1942). Efficient algorithm, x-ray crystallography.

Cooley-Tukey (1965). Monitoring nuclear tests in Soviet Union andtracking submarines. Rediscovered and popularized FFT.

Importance not fully realized until advent of digital computers.

31

Polynomials: Coefficient Representation

Polynomial. [coefficient representation]

Add. O(n) arithmetic operations.

Evaluate. O(n) using Horner's method.

Multiply (convolve). O(n2) using brute force.

!

A(x) = a0 + a1x + a2x2

+L+ an"1x

n"1

!

B(x) = b0 +b1x +b2x2

+L+ bn"1x

n"1

!

A(x)+ B(x) = (a0 +b0 )+ (a1 +b1)x +L+ (an"1 +b

n"1)xn"1

!

A(x) = a0 + (x (a1 + x (a2 +L+ x (an"2 + x (an"1))L))

!

A(x)" B(x) = ci xi

i =0

2n#2

$ , where ci = a j bi# j

j =0

i

$

32

A Modest PhD Dissertation Title

"New Proof of the Theorem That Every Algebraic RationalIntegral Function In One Variable can be Resolved intoReal Factors of the First or the Second Degree."

- PhD dissertation, 1799 the University of Helmstedt

33

Polynomials: Point-Value Representation

Fundamental theorem of algebra. [Gauss, PhD thesis] A degree npolynomial with complex coefficients has exactly n complex roots.

Corollary. A degree n-1 polynomial A(x) is uniquely specified by itsevaluation at n distinct values of x.

x

y

xj

yj = A(xj )

34

Polynomials: Point-Value Representation

Polynomial. [point-value representation]

Add. O(n) arithmetic operations.

Multiply (convolve). O(n), but need 2n-1 points.

Evaluate. O(n2) using Lagrange's formula.

!

A(x) : (x0, y0 ), K, (xn-1, yn"1)

B(x) : (x0, z0 ), K, (xn-1, zn"1)

!

A(x)+B(x) : (x0, y0 + z0 ), K, (xn-1, yn"1 + zn"1)

!

A(x) = yk

(x " x j )j#k

$

(xk " x j )j#k

$k=0

n"1

%

!

A(x) " B(x) : (x0, y0 " z0 ), K, (x2n-1, y2n#1" z2n#1)

35

Converting Between Two Polynomial Representations

Tradeoff. Fast evaluation or fast multiplication. We want both!

Goal. Efficient conversion between two representations ⇒ all ops fast.

coefficient

representation

O(n2)

multiply

O(n)

evaluate

point-value O(n) O(n2)

!

a0, a1, ..., an-1

!

(x0, y0 ), K, (xn"1, yn"1)

coefficient representation point-value representation

36

Converting Between Two Representations: Brute Force

Coefficient ⇒ point-value. Given a polynomial a0 + a1 x + ... + an-1 xn-1,evaluate it at n distinct points x0 , ..., xn-1.

Running time. O(n2) for matrix-vector multiply (or n Horner's).

!

y0

y1

y2

M

yn"1

#

$

% % % % % %

&

'

( ( ( ( ( (

=

1 x0

x0

2L x

0

n"1

1 x1

x1

2L x

1

n"1

1 x2

x2

2L x

2

n"1

M M M O M

1 xn"1 xn"12

L xn"1n"1

#

$

% % % % % %

&

'

( ( ( ( ( (

a0

a1

a2

M

an"1

#

$

% % % % % %

&

'

( ( ( ( ( (

37

Converting Between Two Representations: Brute Force

Point-value ⇒ coefficient. Given n distinct points x0, ... , xn-1 and valuesy0, ... , yn-1, find unique polynomial a0 + a1x + ... + an-1 xn-1, that has givenvalues at given points.

Running time. O(n3) for Gaussian elimination.

!

y0

y1

y2

M

yn"1

#

$

% % % % % %

&

'

( ( ( ( ( (

=

1 x0

x0

2L x

0

n"1

1 x1

x1

2L x

1

n"1

1 x2

x2

2L x

2

n"1

M M M O M

1 xn"1 xn"12

L xn"1n"1

#

$

% % % % % %

&

'

( ( ( ( ( (

a0

a1

a2

M

an"1

#

$

% % % % % %

&

'

( ( ( ( ( (

Vandermonde matrix is invertible iff xi distinct

or O(n2.376) via fast matrix multiplication

38

Divide-and-Conquer

Decimation in frequency. Break up polynomial into low and high powers. A(x) = a0 + a1x + a2x2 + a3x3 + a4x4 + a5x5 + a6x6 + a7x7. Alow(x) = a0 + a1x + a2x2 + a3x3. Ahigh (x) = a4 + a5x + a6x2 + a7x3. A(x) = Alow(x) + x4 Ahigh(x).

Decimation in time. Break polynomial up into even and odd powers. A(x) = a0 + a1x + a2x2 + a3x3 + a4x4 + a5x5 + a6x6 + a7x7. Aeven(x) = a0 + a2x + a4x2 + a6x3. Aodd (x) = a1 + a3x + a5x2 + a7x3. A(x) = Aeven(x2) + x Aodd(x2).

39

Coefficient to Point-Value Representation: Intuition

Coefficient ⇒ point-value. Given a polynomial a0 + a1x + ... + an-1 xn-1,evaluate it at n distinct points x0 , ..., xn-1.

Divide. Break polynomial up into even and odd powers. A(x) = a0 + a1x + a2x2 + a3x3 + a4x4 + a5x5 + a6x6 + a7x7. Aeven(x) = a0 + a2x + a4x2 + a6x3. Aodd (x) = a1 + a3x + a5x2 + a7x3. A(x) = Aeven(x2) + x Aodd(x2). A(-x) = Aeven(x2) - x Aodd(x2).

Intuition. Choose two points to be ±1. A( 1) = Aeven(1) + 1 Aodd(1). A(-1) = Aeven(1) - 1 Aodd(1). Can evaluate polynomial of degree ≤ n

at 2 points by evaluating two polynomialsof degree ≤ ½n at 1 point.

we get to choose which ones!

40

Coefficient to Point-Value Representation: Intuition


Divide. Break polynomial up into even and odd powers. A(x) = a0 + a1x + a2x2 + a3x3 + a4x4 + a5x5 + a6x6 + a7x7. Aeven(x) = a0 + a2x + a4x2 + a6x3. Aodd (x) = a1 + a3x + a5x2 + a7x3. A(x) = Aeven(x2) + x Aodd(x2). A(-x) = Aeven(x2) - x Aodd(x2).

Intuition. Choose four complex points to be ±1, ±i. A(1) = Aeven(1) + 1 Aodd(1). A(-1) = Aeven(1) - 1 Aodd(1). A( i ) = Aeven(-1) + i Aodd(-1). A( -i ) = Aeven(-1) - i Aodd(-1).

Can evaluate polynomial of degree ≤ nat 4 points by evaluating two polynomialsof degree ≤ ½n at 2 points.

we get to choose which ones!

41

Discrete Fourier Transform


Key idea. Choose xk = ωk where ω is principal nth root of unity.

DFT

!

y0

y1

y2

y3

M

yn"1

#

$

% % % % % % %

&

'

( ( ( ( ( ( (

=

1 1 1 1 L 1

1 )1 )2 )3L )n"1

1 )2 )4 )6L )2(n"1)

1 )3 )6 )9L )3(n"1)

M M M M O M

1 )n"1 )2(n"1) )3(n"1)L )(n"1)(n"1)

#

$

% % % % % % %

&

'

( ( ( ( ( ( (

a0

a1

a2

a3

M

an"1

#

$

% % % % % % %

&

'

( ( ( ( ( ( (

Fourier matrix Fn

42

Roots of Unity

Def. An nth root of unity is a complex number x such that xn = 1.

Fact. The nth roots of unity are: ω0, ω1, …, ωn-1 where ω = e 2π i / n.Pf. (ωk)n = (e 2π i k / n) n = (e π i ) 2k = (-1) 2k = 1.

Fact. The ½nth roots of unity are: ν0, ν1, …, νn/2-1 where ν = ω2 = e 4π i / n.

ω0 = ν0 = 1

ω1

ω2 = ν1 = i

ω3

ω4 = ν2 = -1

ω5

ω6 = ν3 = -i

ω7

n = 8

43

Fast Fourier Transform

Goal. Evaluate a degree n-1 polynomial A(x) = a0 + ... + an-1 xn-1 at itsnth roots of unity: ω0, ω1, …, ωn-1.

Divide. Break up polynomial into even and odd powers. Aeven(x) = a0 + a2x + a4x2 + … + an-2 x n/2 - 1. Aodd (x) = a1 + a3x + a5x2 + … + an-1 x n/2 - 1. A(x) = Aeven(x2) + x Aodd(x2).

Conquer. Evaluate Aeven(x) and Aodd(x) at the ½nth

roots of unity: ν0, ν1, …, νn/2-1.

Combine. A(ω k) = Aeven(ν k) + ω k Aodd (ν k), 0 ≤ k < n/2 A(ω k+ ½n) = Aeven(ν k) – ω k Aodd (ν k), 0 ≤ k < n/2

ωk+ ½n = -ωk

νk = (ωk )2

νk = (ωk + ½n )2

44

fft(n, a0,a1,…,an-1) { if (n == 1) return a0

(e0,e1,…,en/2-1) ← FFT(n/2, a0,a2,a4,…,an-2) (d0,d1,…,dn/2-1) ← FFT(n/2, a1,a3,a5,…,an-1)

for k = 0 to n/2 - 1 { ωk ← e2πik/n

yk+n/2 ← ek + ωk dk yk+n/2 ← ek - ωk dk }

return (y0,y1,…,yn-1)}

FFT Algorithm

45

FFT Summary

Theorem. FFT algorithm evaluates a degree n-1 polynomial at each ofthe nth roots of unity in O(n log n) steps.

Running time.

!

a0, a1, ..., an-1

!

("0, y0 ), ..., ("

n#1, yn#1)

O(n log n)

coefficientrepresentation

point-valuerepresentation

!

T (n) = 2T (n /2) + "(n) # T (n) = "(n logn)

???

assumes n is a power of 2

46

Recursion Tree

a0, a1, a2, a3, a4, a5, a6, a7

a1, a3, a5, a7a0, a2, a4, a6

a3, a7a1, a5a0, a4 a2, a6

a0 a4 a2 a6 a1 a5 a3 a7

"bit-reversed" order

000 100 010 110 001 101 011 111

perfect shuffle

47

Inverse Discrete Fourier Transform

Point-value ⇒ coefficient. Given n distinct points x0, ... , xn-1 and valuesy0, ... , yn-1, find unique polynomial a0 + a1x + ... + an-1 xn-1, that has givenvalues at given points.

Inverse DFT

!

a0

a1

a2

a3

M

an"1

#

$

% % % % % % %

&

'

( ( ( ( ( ( (

=

1 1 1 1 L 1

1 )1 )2 )3L )n"1

1 )2 )4 )6L )2(n"1)

1 )3 )6 )9L )3(n"1)

M M M M O M

1 )n"1 )2(n"1) )3(n"1)L )(n"1)(n"1)

#

$

% % % % % % %

&

'

( ( ( ( ( ( (

"1

y0

y1

y2

y3

M

yn"1

#

$

% % % % % % %

&

'

( ( ( ( ( ( (

Fourier matrix inverse (Fn) -1

48

Claim. Inverse of Fourier matrix Fn is given by following formula.

Consequence. To compute inverse FFT, apply same algorithm but use ω-1 = e -2π i / n as principal nth root of unity (and divide by n).

!

Gn

=1

n

1 1 1 1 L 1

1 "#1 "#2 "#3L "#(n#1)

1 "#2 "#4 "#6L "#2(n#1)

1 "#3 "#6 "#9L "#3(n#1)

M M M M O M

1 "#(n#1) "#2(n#1) "#3(n#1)L "#(n#1)(n#1)

$

%

& & & & & & &

'

(

) ) ) ) ) ) )

Inverse DFT

!

1

nFn

is unitary

49

Inverse FFT: Proof of Correctness

Claim. Fn and Gn are inverses.Pf.

Summation lemma. Let ω be a principal nth root of unity. Then

Pf. If k is a multiple of n then ωk = 1 ⇒ series sums to n. Each nth root of unity ωk is a root of xn - 1 = (x - 1) (1 + x + x2 + ... + xn-1). if ωk ≠ 1 we have: 1 + ωk + ωk(2) + … + ωk(n-1) = 0 ⇒ series sums to 0. ▪

!

" k j

j=0

n#1

$ =n if k % 0 mod n

0 otherwise

& ' (

!

Fn Gn( ) k " k = 1

n#k j #$ j " k

j=0

n$1

% = 1

n#(k$ " k ) j

j=0

n$1

% = 1 if k = " k

0 otherwise

& ' (

summation lemma

50

Inverse FFT: Algorithm

ifft(n, a0,a1,…,an-1) { if (n == 1) return a0

(e0,e1,…,en/2-1) ← FFT(n/2, a0,a2,a4,…,an-2) (d0,d1,…,dn/2-1) ← FFT(n/2, a1,a3,a5,…,an-1)

for k = 0 to n/2 - 1 { ωk ← e-2πik/n

yk+n/2 ← (ek + ωk dk) / n yk+n/2 ← (ek - ωk dk) / n }

return (y0,y1,…,yn-1)}

51

Inverse FFT Summary

Theorem. Inverse FFT algorithm interpolates a degree n-1 polynomialgiven values at each of the nth roots of unity in O(n log n) steps.

assumes n is a power of 2

!

a0, a1,K, an-1

!

("0, y0 ), K, ("

n#1, yn#1)

O(n log n)

coefficientrepresentation

O(n log n) point-valuerepresentation

52

Polynomial Multiplication

Theorem. Can multiply two degree n-1 polynomials in O(n log n) steps.

!

a0, a1,K, an-1

b0, b1,K, bn-1

!

c0, c1,K, c2n-2

!

A("0), ..., A("

2n#1)

B(" 0), ..., B(" 2n#1

)

!

C("0), ..., C("

2n#1)

O(n)

point-value multiplication

O(n log n)2 FFTs inverse FFT O(n log n)

coefficientrepresentation coefficient

representation

pad with 0s to make n a power of 2

53

FFT in Practice ?

54

FFT in Practice

Fastest Fourier transform in the West. [Frigo and Johnson] Optimized C library. Features: DFT, DCT, real, complex, any size, any dimension. Won 1999 Wilkinson Prize for Numerical Software. Portable, competitive with vendor-tuned code.

Implementation details. Instead of executing predetermined algorithm, it evaluates your

hardware and uses a special-purpose compiler to generate anoptimized algorithm catered to "shape" of the problem.

Core algorithm is nonrecursive version of Cooley-Tukey. O(n log n), even for prime sizes.

Reference: http://www.fftw.org

55

Integer Multiplication, Redux

Integer multiplication. Given two n bit integers a = an-1 … a1a0 andb = bn-1 … b1b0, compute their product a ⋅ b.

Convolution algorithm. Form two polynomials. Note: a = A(2), b = B(2). Compute C(x) = A(x) ⋅ B(x). Evaluate C(2) = a ⋅ b. Running time: O(n log n) complex arithmetic operations.

Theory. [Schönhage-Strassen 1971] O(n log n log log n) bit operations.Theory. [Fürer 2007] O(n log n 2O(log *n)) bit operations.

!

A(x) = a0 + a1x + a2x2

+L+ an"1x

n"1

!

B(x) = b0 +b1x +b2x2

+L+ bn"1x

n"1

56

Integer Multiplication, Redux

Integer multiplication. Given two n bit integers a = an-1 … a1a0 andb = bn-1 … b1b0, compute their product a ⋅ b.

Practice. [GNU Multiple Precision Arithmetic Library]It uses brute force, Karatsuba, and FFT, depending on the size of n.

"the fastest bignum library on the planet"

57

Integer Arithmetic

Fundamental open question. What is complexity of arithmetic?

addition

Operation

O(n)

Upper Bound

Ω(n)

Lower Bound

multiplication O(n log n 2O(log*n)) Ω(n)

division O(n log n 2O(log*n)) Ω(n)

58

Factoring

Factoring. Given an n-bit integer, find its prime factorization.

2773 = 47 × 59

267-1 = 147573952589676412927 = 193707721 × 761838257287

RSA-704($30,000 prize if you can factor)

74037563479561712828046796097429573142593188889231289084936232638972765034028266276891996419625117843995894330502127585370118968098286733173273108930900552505116877063299072396380786710086096962537934650563796359

a disproof of Mersenne's conjecture that 267 - 1 is prime

59

Factoring and RSA

Primality. Given an n-bit integer, is it prime?Factoring. Given an n-bit integer, find its prime factorization.

Significance. Efficient primality testing ⇒ can implement RSA.Significance. Efficient factoring ⇒ can break RSA.

Theorem. [AKS 2002] Poly-time algorithm for primality testing.

60

Shor's Algorithm

Shor's algorithm. Can factor an n-bit integer in O(n3) time on aquantum computer.

Ramification. At least one of the following is wrong: RSA is secure. Textbook quantum mechanics. Extending Church-Turing thesis.

algorithm uses quantum QFT !

61

Shor's Factoring Algorithm

Period finding.

Theorem. [Euler] Let p and q be prime, and let N = p q. Then, thefollowing sequence repeats with a period divisible by (p-1) (q-1):

Consequence. If we can learn something about the period of thesequence, we can learn something about the divisors of (p-1) (q-1).

by using random values of x, we get the divisors of (p-1) (q-1),and from this, can get the divisors of N = p q

1 2 4 8 16 32 64 128 …2 i

1 2 4 8 1 2 4 8 …2 i mod 15

1 2 4 8 16 11 1 2 …2 i mod 21period = 4

period = 6

x mod N, x2 mod N, x3 mod N, x4 mod N, …

Extra Slides

63

Fourier Matrix Decomposition

!

y = Fn a = In /2

Dn /2

In /2"Dn /2

#

$ %

&

' (

Fn /2aeven

Fn /2aodd

#

$ %

&

' (

!

I4

=

1 0 0 0

0 1 0 0

0 0 1 0

0 0 0 1

"

#

$ $ $ $

%

&

' ' ' '

!

D4

=

"00 0 0

0 "10 0

0 0 "20

0 0 0 "3

#

$

% % % %

&

'

( ( ( (

!

Fn

=

1 1 1 1 L 1

1 "1 "2 "3L "n#1

1 "2 "4 "6L "2(n#1)

1 "3 "6 "9L "3(n#1)

M M M M O M

1 "n#1 "2(n#1) "3(n#1)L "(n#1)(n#1)

$

%

& & & & & & &

'

(

) ) ) ) ) ) )

!

a =

a0

a1

a2

a3

"

#

$ $ $ $

%

&

' ' ' '

how to multiply · matrix multiplication. given two n-by-n matrices a and b, compute c = ab....

Documents