- 50 Views
- Uploaded on
- Presentation posted in: General

Some recent results in mathematics related to data transmission:

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Some recent results in mathematics related to data transmission:

Michel Waldschmidt

Université P. et M. Curie - Paris VI

Centre International de Mathématiques Pures et Appliquées - CIMPA

India, October-November 2007

http://www.math.jussieu.fr/~miw/

French Science Today

Some recent results in mathematics related to data transmission

India October- November 2007

Starting with card tricks, we show how mathematical tools are used to detect and to correct errors occuring in the transmission of data.

These so-called "error-detecting codes" and "error-correcting codes" enable identification and correction of the errors caused by noise or other impairments during transmission from the transmitter to the receiver. They are used in compact disks to correct errors caused by scratches, in satellite broadcasting, in digital money transfers, in telephone connexions, they are useful for improving the reliability of data storage media as well as to correct errors cause when a hard drive fails. The National Aeronautics and Space Administration (NASA) has used many different error-correcting codes for deep-space telecommunications and orbital missions.

http://www.math.jussieu.fr/~miw/

French Science Today

India November 2007

Some recent results in mathematics related to data transmission

Most of the theory arises from earlier developments of mathematics which were far removed from any concrete application. One of the main tools is the theory of finite fields, which was invented by Galois in the XIXth century, for solving polynomial equations by means of radicals. The first error-correcting code happened to occur in a sport newspaper in Finland in 1930. The mathematical theory of information was created half a century ago by Claude Shannon. The mathematics behind these technical devices are being developped in a number of scientific centers all around the world, including in India and in France.

http://www.math.jussieu.fr/~miw/

French Science Today

Mathematical aspects of

Coding Theory in France:

The main teams in the domain are gathered in the group

C2 ''Coding Theory and Cryptography'' ,

which belongs to a more general group (GDR) ''Mathematical Informatics''.

http://www.math.jussieu.fr/~miw/

French Science Today

The most important are:

INRIA Rocquencourt

Université de Bordeaux

ENST Télécom Bretagne

Université de Limoges

Université de Marseille

Université de Toulon

Université de Toulouse

http://www.math.jussieu.fr/~miw/

INRIA

Brest

Limoges

Bordeaux

Marseille

Toulon

Toulouse

http://www-rocq.inria.fr/codes/

National Research Institute in Computer Science and Automatic

http://www.math.u-bordeaux1.fr/maths/

Lattices

and

combinatorics

http://departements.enst-bretagne.fr/sc/recherche/turbo/

Turbocodes

http://www.xlim.fr/

Arithmetic and Information Theory

Algebraic geometry over finite fields

http://grim.univ-tln.fr/

Université du Sud Toulon-Var

Boolean functions

Université de Toulouse

Le Mirail

Algebraic geometry over finite fields

http://www.univ-tlse2.fr/grimm/algo

http://www.gdr-im.fr/

- The GDR ''Mathematical Informatics'' gathers all the french teams which work on computer science problems with mathematical methods.

http://www.gdr-im.fr/

- Calcul Formel (Symbolic computation)
- ARITH: Arithmétique (Arithmetics)
- COMBALG : Combinatoire algébrique (Algebraic Combinatorics)

French Science Today

Mathematical Aspects of

Coding Theory in India:

Indian Institute of Technology Bombay

Indian Institute of Science Bangalore

Indian Institute of Technology Kanpur

Panjab University Chandigarh

University of Delhi Delhi

Chandigarh

Delhi

Kanpur

Bombay

Bangalore

http://www.iitb.ac.in/

- Department of Mathematics
- Department of Electrical Engineering

http://www.iisc.ernet.in/

- Finite fields and Coding Theory classification of permutation polynomials, study of PAPR of families of codes, construction of codes with low PAPR.

- Department of Mathematics

peak-to-average power

http://www.iitk.ac.in/

http://www.puchd.ac.in/

http://www.du.ac.in/

Department of Mathematics

http://www.ias.ac.in/resonance/

- How Numbers Protect Themselves
- The Hamming Codes Volume 2 Number 1
- Reed Solomon Codes Volume 2 Number 3

Playing cards

- Among a collection of playing cards, you select one without telling me which one it is.
- I ask you some questions where you answer yes or no.
- Then I am able to tell you which card you selected.

- You select one of these two cards
- I ask you one question and you answer yes or no.
- I am able to tell you which card you selected.

- Question: is-it this one?

First question: is-it one of these two?

Y Y

Y N

N Y

N N

YYY

YYN

YNY

YNN

NYY

NYN

NNY

NNN

- 0 / 1
- Yin — / Yang - -
- True / False
- White / Black
- + / -
- Heads / Tails (tossing or flipping a coin)

0 0 0

0

0 0 1

1

0 1 0

2

0 1 1

3

1 0 0

4

1 0 1

5

1 1 0

6

1 1 1

7

Add one question =

multiply the number of cards by 2

- If you select one card among a set of 16, I shall know which one it is, once you answer my 4 questions by yes or no.

12

0

4

8

13

1

5

9

10

14

6

2

11

15

7

3

0 0 0 0

1 1 0 0

0 1 0 0

1 0 0 0

0 0 0 1

1 1 0 1

0 1 0 1

1 0 0 1

0 0 1 0

1 1 1 0

0 1 1 0

1 0 1 0

1 1 1 1

1 0 1 1

0 1 1 1

0 0 1 1

Y Y Y Y

N N Y Y

Y N Y Y

N Y Y Y

Y Y Y N

N N Y N

Y N Y N

N Y Y N

Y Y N Y

N N N Y

Y N N Y

N Y N Y

N N N N

N Y N N

Y N N N

Y Y N N

0000

1100

0100

1000

1101

0001

0101

1001

0110

1010

1110

0010

0111

1111

0011

1011

- Is the first digit 0 ?
- Is the second digit 0?
- Is the third digit 0?
- Is the fourth digit 0?

More difficult:

One answer may be wrong!

- Consider the same problem, but you are allowed to give (at most) one wrong answer.
- How many questions are required so that I am able to know whether your answers are right or not? And if they are right, to know the card you selected?

- If I ask one more question, I shall be able to detect if there is one of your answers which is not compatible with the others.
- And if you made no mistake, I shall tell you which is the card you selected.

- With two cards I just repeat twice the same question.
- If both your answers are the same, you did not lie and I know which card you selected
- If your answers are not the same, I know that one is right and one is wrong (but I don’t know which one is correct!).

First question: is-it one of these two?

Third question: is-it one of these two?

Y Y Y

Y N N

N Y N

N N Y

0 0 0

0 1 1

1 0 1

1 1 0

0 0 0

0 0 1

0 1 1

0 1 0

1 0 1

1 0 0

1 1 0

1 1 1

Correct triple of answers:

Wrong triple of answers

One change in a correct triple of answers

yields a wrong triple of answers

- even + even = even
- even + odd = odd
- odd + even = odd
- odd + odd = even

- 0 + 0 = 0
- 0 + 1 = 1
- 1 + 0 = 1
- 1 + 1 = 0

- Use one more bit which is the Boolean sum of the previous ones.
- Now for a correct answer the sum of the bits should be 0.
- If there is exactly one error, the parity bit will detect it: the sum of the bits will be 1 instead of 0.

YYYY

0000

YYNN

0011

YNYN

0101

YNNY

0110

NYYN

1001

NYNY

1010

NNYY

1100

NNNN

1111

Use the 3 previous questions plus the parity bit question

YYYYY

NNYYY

YNYYN

NYYYN

YYYNN

NNYNN

YNYNY

NYYNY

YYNYN

NNNYN

YNNYY

NYNYY

NNNNY

NYNNN

YNNNN

YYNNY

- Again I ask you questions where your answer is yes or no, again you are allowed to give at most one wrong answer, but now I want to be able to know which card you selected - and also to tell you whether and when you lied.

- I repeat the same question three times.
- The most frequent answer is the right one: vote with the majority.
- 2 cards, 3 questions, corrects 1 error.

If I repeat my two questions three times each, I need 6 questions

Better way: repeat each of the two questions twice only, and use the parity check bit.

5 questions suffice

- 0 0 0 0 0or Y Y Y Y Y
- 0 1 0 1 1or Y N Y N N
- 1 0 1 0 1or N Y N Y N
- 1 1 1 1 0or N N N N Y
- If there is just one mistake, it is easy to correct it.
- 4 cards, 5 questions, corrects 1 error.

If I repeat 3 times

my 3 questions,

I need 9 questions

With 6 questions

only I can correct

one error

- Ask the three questions giving the right answer if there is no error, then use the parity check for questions (1,2), (1,3) and (2,3).
- Right answers :
(a, b, c, a+b, a+c, b+c)

with a, b, c replaced by 0 or 1

Error correcting codes

- Coding theory is the branch of mathematics concerned with transmitting data across noisy channels and recovering the message. Coding theory is about making messages easy to read: don't confuse it with cryptography which is the art of making messages hard to read!

- In 1948, Claude Shannon, working at Bell Laboratories in the USA, inaugurated the whole subject of coding theory by showing that it was possible to encode messages in such a way that the number of extra bits transmitted was as small as possible. Unfortunately his proof did not give any explicit recipes for these optimal codes.

Around the same time, Richard Hamming, also at Bell Labs, was using machines with lamps and relays having an error detecting code. The digits from 1 to 9 were send on ramps of 5 lamps with two on and three out. There were very frequent errors which were easy to detect and then one had to restart the process.

- For his researches, Hamming was allowed to have the machine working during the week-end only, and they were on the automatic mode. At each error the machine stopped until the next monday morning.
- "If it can detect the error," complained Hamming, "why can't it correct it! "

- He decided to find a device so that the machine would not only detect the errors but also correct them.
- In 1950, he published details of his work on explicit error-correcting codes with information transmission rates more efficient than simple repetition.
- His first attempt produced a code in which four data bits were followed by three check bits which allowed not only the detection, but also the correction of a single error.

- 1949: Marcel Golay (specialist of radars): produced two remarkably efficient codes.
- Eruptions on Io (Jupiter’s volcanic moon)
- 1963 John Leech: uses Golay’s ideas for sphere packing in dimension 24 - classification of finite simple groups
- 1971: no other perfect code than the two found by Golay.

- Telephone
- CD or DVD
- Image transmission
- Sending information through the Internet
- Radio control of satellites

- Transmitions by satellites
- Compact discs
- Cellular phones

Olympus Mons on Mars Planet

Image from Mariner 2 in 1971.

- Between 1969 and 1973 the NASA Mariner probes used a powerful Reed-Muller code capable of correcting 7 errors out of 32 bits transmitted, consisting now of 6 data bits and 26 check bits! Over 16,000 bits per second were relayed back to Earth.

The North polar cap of Mars,

taken by Mariner 9 in 1972.

- Journey: Cape Canaveral, Jupiter, Saturn, Uranus, Neptune.
- Sent information by means of a binary code which corrected 3 errors on words of length 24.

- Sent black and white photographs of Mars
- Grid of 600 by 600, each pixel being assigned one of 64 brightness levels
- Reed-Muller code with 64 words of 32 letters, minimal distance 16, correcting 7 errors, rate 3/16

- Color photos of Jupiter and Saturn
- Golay code with 4096=212 words of 24 letters, minimal distance 8, corrects 3 errors, rate 1/2.
- 1998: lost of control of Soho satellite recovered thanks to double correction by turbo code.

- The power of the radio transmitters on these craft is only a few watts, yet this information is reliably transmitted across hundreds of millions of miles without being completely swamped by noise.

Sojourner rover

and Mars Pathfinder lander

- On a CD as well as on a computer, each sound is coded by a sequence of 0’s and 1’s, grouped in octets
- Further octets are added which detect and correct small mistakes.
- In a CD, two codes join forces and manage to handle situations with vast number of errors.

On CDs the signal in encoded digitally.

To guard against scratches, cracks and similar

damage, two "interleaved" codes which can correct

up to 4,000 consecutive errors (about 2.5 mm of track)

are used.

Using a finite field with 256 elements, it is possible to correct 2 errors in each word of 32 octets with 4 control octets for 28 information octets.

- After processing of the signal in the CD player, these errors do not lead to any disturbing noise.
- Without error-correcting codes, there would be no CD.

- 1980’s, agreement between Sony and Philips: norm for storage of data on audio CD’s.
- 44 100 times per second, 16 bits in each of the two stereo channels

- Algebra
(discrete mathematics finite fields, linear algebra,…)

- Geometry
- Probability and statistics

- Solving algebraic equations with radicals: Finite fields theory Evariste Galois(1811-1832)
- Construction of regular polygons with rule and compass
- Group theory

Srinivasa Ramanujan (1887-1920)

Only certain words are allowed (code = dictionary of valid words).

The « useful » letters (data bits) carry the information, the other ones (control or check bits) allow detecting or correcting errors.

Send twice each bit

2 code words among 4=22possible words

(1 data bit, 1 check bit)

Code words

(two letters)

0 0

and

1 1

Rate: 1/2

Principle of codes detecting one error:

Two distinct code words

have at least two distinct letters

Code words (three letters):

0 0 0

0 1 1

1 0 1

1 1 0

Parity bit : (x yz) with z=x+y.

4 code words (among 8words with 3 letters),

2 data bits, 1 check bit.

Rate: 2/3

2

0 0 00 0 1

0 1 1 0 1 0

1 0 11 0 0

1 1 0 1 1 1

Two distinct code words

have at least two distinct letters.

2

- In the International Standard Book Number (ISBN) system used to identify books, the last of the ten-digit number is a check bit.
- The Chemical Abstracts Service (CAS) method of identifying chemical compounds, the United States Postal Service (USPS) use check digits.
- Modems, computer memory chips compute checksums.
- One or more check digits are commonly embedded in credit card numbers.

Send each bit three times

2 code words

among 8possible ones

(1 data bit, 2 check bits)

Code words

(three letters)

0 0 0

1 1 1

Rate: 1/3

- Correct 0 0 1 as 0 0 0
- 0 1 0 as 0 0 0
- 1 0 0 as 0 0 0
and

- 1 1 0 as 1 1 1
- 1 0 1 as 1 1 1
- 0 1 1 as 1 1 1

Principle of codes correcting one error:

Two distinct code words have at least three distinct letters

= number of places in which the two words

differ

Examples:

(0,0,1) and (0,0,0) have distance 1

(1,0,1) and (1,1,0) have distance 2

(0,0,1) and (1,1,0) have distance 3

Richard W. Hamming (1915-1998)

- The unit sphere around a word includes the words at distance at most 1

The code (0 0 0) (1 1 1)

- The set of words with three letters (eight elements) splits into two balls
- The centers are (0,0,0) and (1,1,1)
- Each of the two balls consists of elements at distance at most 1 from the center

Two or three 0

Two or three 1

(0,0,1)

(1,0,1)

(0,1,0)

(1,1,0)

(0,0,0)

(1,1,1)

(1,0,0)

(0,1,1)

- Code words:a b a b a+b
0 0 0 0 0

0 1 0 1 1

1 0 1 0 1

1 1 1 1 0

- Two code words have distance at least 3
Rate : 2/5.

- Code words:a b c a+b a+c b+c
0 0 0 0 0 0 1 0 0 1 1 0

0 0 1 0 1 1 1 0 1 1 0 1

0 1 0 1 0 1 1 1 0 0 1 1

0 1 1 1 1 0 1 1 1 0 0 0

- Two code words have distance at least 3
Rate : 1/2.

Hamming’s code, 1950

Rate : 4/7.

Generalization of the

parity check bit

How to compute e , f , g

from a , b , c , d.

e=a+b+d

d

a

b

f=a+c+d

c

g=a+b+c

Words of 7 letters

Code words: (16=24 among 128=27)

(a b c de f g)

with

e=a+b+d

f=a+c+d

g=a+b+c

Rate: 4/7

0 0 0 0 0 0 0

0 0 0 1 1 1 0

0 0 1 0 0 1 1

0 0 1 1 1 0 1

0 1 0 0 1 0 1

0 1 0 1 0 1 1

0 1 1 0 1 1 0

0 1 1 1 0 0 0

1 0 0 0 1 1 1

1 0 0 1 0 0 1

1 0 1 0 1 0 0

1 0 1 1 0 1 0

1 1 0 0 0 1 0

1 1 0 1 1 0 0

1 1 1 0 0 0 1

1 1 1 1 1 1 1

Two distinct code words have at least three distinct letters

It is a linear code (the sum of two code words is a code word) and the 16 balls of radius 1 with centers in the code words cover all the space of the 128 binary words of length 7

(each word has 7 neighbors (7+1)16= 256).

Playing cards

Replace the cards by labels from 0 to 15 and write the binary expansions of these:

0000, 0001, 0010, 0011

0100, 0101, 0110, 0111

1000, 1001, 1010, 1011

1100, 1101, 1110, 1111

Using the Hamming code, get 7 digits.

Select the questions so that Yes=0 and No=1

- Is the first binary digit 1?
- Is the second binary digit 1?
- Is the third binary digit 1?
- Is the fourth binary digit 1?
- Is the number in {1,2,4,7,9,10,12,15}?
- Is the number in {1,2,5,6,8,11,12,15}?
- Is the number in {1,3,4,6,8,10,13,15}?

The Hat Problem

- Three people are in a room, each has a hat on his head, the color of which is black or white. Hat colors are chosen randomly. Everybody sees the color of the hat on everyone’s head, but not on their own. People do not communicate with each other.
- Everyone gets to guess (by writing on a piece of paper) the color of their hat. They may write: Black/White/Abstain.

- The people in the room win together or lose together.
- The team wins if at least one of the three people did not abstain, and everyone who did not abstain guessed the color of their hat correctly.
- How will this team decide a good strategy with a high probability of winning?

- Simple strategy: they agree that two of them abstain and the other guesses randomly.
- Probability of winning: 1/2.
- Is it possible to do better?

- Hint:
Improve the odds by using the available information: everybody sees the color of the hat on everyone’s head but himself.

- Better strategy: if a member sees two different colors, he abstains. If he sees the same color twice, he guesses that his hat has the other color.

The two people with white hats see one white hat and one black hat, so they abstain.

The one with a black hat sees two white hats, so he writes black.

They win!

The two people with black hats see one white hat and one black hat, so they abstain.

The one with a white hat sees two black hats, so he writes white.

They win!

Everybody sees two white hats, and therefore writes black on the paper.

They lose

Everybody sees two black hats, and therefore writes white on the paper.

They lose

Winning:

two white

or

two black

Losing:

three white

or

three black

- The team wins exactly when the three hats do not have all the same color, that is in 6 cases out of a total of 8
- Probability of winning: 3/4.

- Replace white by 0 and black by 1
hence the distribution of colors becomes a word of three letters on the alphabet {0 , 1}

- Consider the centers of the balls (0,0,0)and(1,1,1).
- The team bets that the distribution of colors is not one of the two centers.

- One color occurs exactly twice (the word has both digits 0and 1).
- Exactly one member of the team sees twice the same color: this corresponds to 00in case he sees two white hats, 1 1in case he sees two black hats.
- Hence he knows the center of the ball: (0 , 0 , 0) in the first case, (1, 1, 1) in the second case.
- He bets the missing digit does not yield the center.

- The two others see two different colors, hence they do not know the center of the ball. They abstain.
- Therefore the team wins when the distribution of colors does not correspond to one of the centers of the balls.
- This is why the team wins in6cases.

- Now if the word corresponding to the distribution of the hats is one of the centers, all members of the team bet the wrong answer!
- They lose in2cases.

For 7 people in the room in place of 3,

which is the best strategy

and its probability of winning?

Answer:

the best strategy gives a

probability of winning of 7/8

- The team bets that the distribution of the hats does not correspond to the 16 elements of the Hamming code
- Loses in 16 cases (they all fail)
- Wins in 128-16=112cases (one bets correctly, the 6others abstain)
- Probability of winning: 112/128=7/8

- A match between two players (or teams) may give three possible results: either player 1 wins, or player 2 wins, or else there is a draw (write 0).
- There is a lottery, and a winning ticket needs to have at least three correct bets. How many tickets should one buy to be sure to win?

- For 4matches, there are 34 = 81 possibilities.
- A bet on4 matches is a sequence of 4symbols {0, 1,2}. Each such ticket has exactly 3 correct answers 8 times.
- Hence each ticket is winning in 9 cases.
- Since 9 9 = 81, a minimum of 9 tickets is required to be sure to win.

0 0 0 01 0 1 22 0 2 1

0 1 1 11 1 2 02 1 0 2

0 2 2 21 2 0 12 2 1 0

Rule: a b a+b a+2b modulo 3

This is an error correcting code on the alphabet

{0, 1, 2} with rate 1/2

- While Shannon and Hamming were working on information transmission in the States, John Leech invented similar codes while working on Group Theory at Cambridge. This research included work on the sphere packing problem and culminated in the remarkable, 24-dimensional Leech lattice, the study of which was a key element in the programme to understand and classify finite symmetry groups.

The kissing number is 12

- Kepler Problem: maximal density of a packing of identical sphères :
p / Ö 18= 0.740 480 49…

Conjectured in 1611.

Proved in 1999 by Thomas Hales.

- Connections with crystallography.

In the past two years the goal of finding explicit codes which reach the limits predicted by Shannon's original work has been achieved. The constructions require techniques from a surprisingly wide range of pure mathematics: linear algebra, the theory of fields and algebraic geometry all play a vital role. Not only has coding theory helped to solve problems of vital importance in the world outside mathematics, it has enriched other branches of mathematics, with new problems as well as new solutions.

- Theoretical questions of existence of specific codes
- connection with cryptography
- lattices and combinatoric designs
- algebraic geometry over finite fields
- equations over finite fields

Some recent results in mathematics related to data transmission:

Michel Waldschmidt

Université P. et M. Curie - Paris VI

Centre International de Mathématiques Pures et Appliquées - CIMPA

India, October-November 2007

http://www.math.jussieu.fr/~miw/