8.8 KiB
CSE5313 Coding and information theory for data science (Lecture 3)
Finite Fields
Why finite fields?
Most information systems are discrete.
- Use bits, byte etc.
Use bits/bytes to represent real numbers.
- Problems of overflow, accuracy, etc.
We wish to build "good" codes \mathcal{C} \subset \mathbb{F}^n:
- Large
\frac{k}{n} - Lage
d_H(\mathcal{C})\implieserror detection/correction, erasure correction.
Idea: Use linear algebraic operations to encode/decode.
F=\mathbb{F}_q, a finite field withqelements.
Finite fields
Fields and field axioms
A field is a set \mathbb{F} with two operations + and \cdot that satisfy the following axioms:
- Associativity:
(a+b)+c = a+(b+c)and(a\cdot b)\cdot c = a\cdot (b\cdot c) - Commutativity:
a+b = b+aanda\cdot b = b\cdot a - Distributivity:
a\cdot (b+c) = a\cdot b + a\cdot c - Existence of Identity elements:
a+0 = aanda\cdot 1 = a - Existence of Inverse elements:
a+(-a) = 0anda\cdot a^{-1} = 1
Every set of elements which satisfies these axioms is a field.
We can "do algebra" over it (matrices, vector spaces, etc.).
Are there finite sets which satisfy the field axioms?
What are the possible sizes of such sets?
Background – Basic number theory
- For
a, b \in \mathbb{N},- Greatest Common Denominator:
\gcd(a, b) =the largest integermsuch thatm|aandm|b. - Lowest Common Multiplier:
\operatorname{lcm}(a, b) =the smallest integermsuch thata|mandb|m.
- Greatest Common Denominator:
a, bare coprime if\gcd(a, b) = 1.- Fact: (Euclid’s lemma) Say
a \geq b,- There exists a quotient
q \geq 0and a remainder0 \leq r < bsuch thata = bq + r.
- There exists a quotient
- Theorem (Euclid): If
\gcd(a, b) = 1then there existm, n \in \mathbb{Z}such thatam + bn = 1.- Proof by repeated application of Euclid’s lemma.
- Example:
- If
a = 3, b = 8, - then
m = -5, n = 2, - satisfy
3 \cdot -5 + 8 \cdot 2 = 1.
- If
Modular arithmetic
Defined a set with addition \oplus and multiplication \odot that satisfy the field axioms.
\mathbb{Z}_p is a field if p is a prime number.
-
Addition and multiplication are defined modulo
p. -
a \oplus b = (a+b) \mod p -
a \odot b = (a\cdot b) \mod p -
0is the additive identity. -
1is the multiplicative identity. -
ahas an additive inversep-a. -
ahas a multiplicative inversea^{-1}such thata \odot a^{-1} = 1.
Proof for existence of multiplicative inverse for a\in \mathbb{Z}_p\setminus \{0\}:
Proof
Since p is prime, \gcd(a, p) = 1.
By euclid's theorem, there exist m, n \in \mathbb{Z} such that am + pn = 1.
Take mod p on both sides:
a_{\mod p}\odot m_{\mod p} \equiv 1_{\mod p}
Thus, m_{\mod p} is the multiplicative inverse of a_{\mod p}.
Polynomials over prime fields is also a field.
(\mathbb{Z}_2,\operatorname{XOR},\operatorname{AND}) is a field.
Polynomials over finite fields
A polynomial over a field \mathbb{Z}_p is a expression of the form:
a(x)=\sum_{i=0}^n a_i x^i
- Polynomial degree: largest index of a non-zero coefficient.
- Polynomial addition:
a(x) \oplus b(x) = \sum_{i=0}^n (a_i \oplus b_i) x^i - Polynomial multiplication:
a(x)\odot b(x) = \sum_{i=0}^n \sum_{j=0}^n a_i \odot b_j x^{i+j} - Polynomial equality:
a(x) = b(x)if and only ifa_i = b_ifor alli. - Polynomial division: suppose
\deg(a(x)) \geq \deg(b(x)), then there exist unique polynomialsq(x)andr(x)such thata(x) = b(x)q(x) \oplus r(x)and\deg(r(x)) < \deg(b(x)). (do long division for polynomials)
denoted as \mathbb{Z}_p[x].
Example
p(x) = x^2 + 6x+3\in \mathbb{Z}_7[x]
p(1) = 1^2 + 6\cdot 1 + 3 = 10 \equiv 3 \mod 7
p(2) = 2^2 + 6\cdot 2 + 3 = 4+5+3 = 12 \equiv 5 \mod 7
Irreducible polynomials
A polynomial p(x) is irreducible if it cannot be factored into two non-constant polynomials.
If \gcd(a(x),b(x))=1, then there exist m(x),n(x)\in \mathbb{Z}_p[x] such that a(x)m(x)\oplus b(x)n(x)=1.
Proved similar to euclid's theorem.
Tip
If a polynomial
p(x)has a root, sayr, thenp(x) = (x-r)q(x)for someq(x)\in \mathbb{Z}_p[x].
Example in \mathbb{Z}_2[x]:
p(x) = x^2 \oplus 1
is reducible because p(x) = (x\oplus 1)(x\oplus 1).
p(x) = x^3 \oplus x \oplus 1
is irreducible.
Proof
We prove by contradiction.
Suppose p(x) is reducible, then p(x) = a(x)b(x) for some a(x),b(x)\in \mathbb{Z}_2[x].
Then \deg(p(x)) = \deg(a(x)) + \deg(b(x)).
Let \deg b(x)=1, then b(x) \in \{x, x\oplus 1\}.
If b(x) = x, then p(0)=0 but p(x) is 1.
If b(x) = x\oplus 1, then p(1)=0 but p(x) is 1.
It is not the case in \mathbb{Z}_2[x], that every polynomial with no root is irreducible. (e.g consider (x^3\oplus x\oplus 1)^2 has no root but is reducible.)
Polynomial modular arithmetic
There exist quotient q(x) and remainder r(x), \deg(r(x)) < \deg(b(x)) such that
a(x) = b(x)q(x) + r(x)
\implies a(x) \mod b(x) = r(x)
"$\mod b(x)$" is an operation on polynomials in \mathbb{Z}_p[x] that:
- Preserves polynomial addition:
a(x) \oplus c(x) \mod b(x) = a(x) \mod b(x) \oplus c(x) \mod b(x)
- Preserves polynomial multiplication:
a(x) \odot c(x) \mod b(x) = a(x) \mod b(x) \odot c(x) \mod b(x)
Extension fields
Let p be a prime number. then (\mathbb{Z}_p[x], \oplus, \odot) is a field.
Fix a polynomial f(x)\in \mathbb{Z}_p[x] of degree t.
Define a set
Elements: polynomials of degree at most t-1 in \mathbb{Z}_p[x]. (finite set, size is p^t.)
Define addition:
a(x) \oplus_f b(x) = (a(x) \oplus b(x)) \mod f(x)
Define multiplication:
a(x) \odot_f b(x) = (a(x) \odot b(x)) \mod f(x)
Denote this set as \mathbb{Z}_p[x] \mod f(x).
This is not a field because it does not have a multiplicative inverse for every element.
Proof
We prove by contradiction.
Suppose there exists a polynomial g(x)\in \mathbb{Z}_p[x] \mod f(x) such that a(x) \odot_f g(x) = 1.
Let p=2,f(x)=x^2\oplus 1.
The polynomials in \mathbb{Z}_2[x] \mod f(x) are \{0, 1, x, x\oplus 1\}.
Consider the modular inverse of (x\oplus 1).
0\odot_f (x\oplus 1) = 01\odot_f (x\oplus 1) = x\oplus 1x\odot_f (x\oplus 1) = (x^2\oplus x)\mod (x^2\oplus 1) = x\oplus 1(x\oplus 1)\odot_f (x\oplus 1) = (x^2\oplus 1)\mod (x^2\oplus 1) = 0
To make our field extension works, we need to find a polynomial f(x) that is irreducible.
Theorem: If f(x) is irreducible over \mathbb{Z}_p, then \mathbb{Z}_p[x] \mod f(x) is a field.
Proof
Let a(x)\in \mathbb{Z}_p[x] \mod f(x), a(x)\neq 0.
Existence of a(x)^{-1} in \mathbb{Z}_p[x] \mod f(x) can be done by Euclid's Theorem.
Since \gcd(a(x),f(x))=1, there exist m(x),n(x)\in \mathbb{Z}_p[x] such that a(x)m(x)\oplus f(x)n(x)=1.
Take mod f(x) on both sides:
a(x)m(x) \mod f(x) = 1 \mod f(x)
Thus, m(x) \mod f(x) is the multiplicative inverse of a(x) \mod f(x).
So a(x)^{-1} = m(x) \mod f(x).
Corollary:
We can extend a prime field \mathbb{Z}_p with irreducible polynomial
Intuitively, we add to \mathbb{Z}_p a new element x that satisfies f(x)=0.
Observation: – We only used the general field properties of \mathbb{Z}_p. – ⇒ any “base field” can be used instead of \mathbb{Z}_p. – ⇒ Any field can be “extended”.
Say we wish to build a field F with 2^8 elements.
-
Option 1:
- Take
\mathbb{Z}_2andf(x)irreducible of degree 8. F = \mathbb{Z}_2[x] \mod f(x).
- Take
-
Option 2:
- Take
\mathbb{Z}_2, andg_1(x) \in \mathbb{Z}_2[x]irreducible of degree 4, F_1 = \mathbb{Z}_2[x] \mod g_1(x). Note|F_1| = 2^4 = 16.- Take
g_2(x) \in F_1[x]irreducible of degree 2. F_2 = F_1[x] \mod g_2(x).
- Take
Uniqueness of the finite field
Theorems:
- As long as it is irreducible, the choice of
f(x)does not matter.- If
f_1(x), f_2(x)are irreducible of the same degree, then\mathbb{Z}_p[x] \mod f_1(x) \cong \mathbb{Z}_p[x] \mod f_2(x).
- If
- Over every
\mathbb{Z}_p(𝑝 prime), there exists an irreducible polynomial of every degree. - All finite fields of the same size are isomorphic.
- All finite fields are of size
p^dfor primepand integerd.
Corollary: This is effectively the only way to construct finite fields!
Extension of fields
\mathbb{R}[x]\mod (x^2+1) is a field, \cong \mathbb{C}.
| Terms | Finite field extension F_1\to F_2 |
\mathbb{R}\to \mathbb{C} |
|---|---|---|
| Base field | any field \mathbb{F}_1 |
\mathbb{R} |
| Irreducible polynomial | f(x) |
x^2+1 |
| New elements added | x |
i |
| Add/mul | \mod f(x) |
\mod (x^2+1) |
You cannot do algebraic extension of \mathbb{Q} to \mathbb{R}.
Transcendental extension: