CMSC 39600: PCPs, codes and inapproximability Oct 2, 2007

Lecture 3: Linearity Testing
Lecturer: Prahladh Harsha Scribe: Joshua A. Grochow
In today’s lecture we’ll go over some preliminaries to the coding theory we will need
throughout the course. In particular, we will go over the Walsh-Hadamard code and the
Blum-Luby-Rubinfeld linearity test. In the following lecture we will use these tools to show
that NP ⊆ PCP[poly, O(1)] (recall that our goal is ultimately to bring the polynomial
amount of randomness down to a logarithmic amount). A common feature throughout the
course is that we will use codes to construct PCPs and vice-versa, construct useful codes
from PCPs.
3.1 Coding Theory – Preliminaries
Coding theory is the study of how to communicate reliably over a noisy channel. The most
common setting is as follows: A message m is put through an encoder E, yielding a value
E(m), also called the codeword (typically much longer than m). The codeword E(m) is
then sent through the noisy channel and arrives at the other end with some noise introduced
η by the channel as E(m) + η. The decoder D then takes E(m) + η as input, and ideally
outputs m as long as the noise is not too much. (Note that it is expected of the decoder
that when applied to E(m) it outputs m.)
Formally, a code is specified by an encoding function C : ¦0, 1¦
k
→ ¦0, 1¦
n
; the outputs
of C are called codewords. The rate of the code C is k/n, i.e. the ratio of input bits to
codeword bits. Heuristically, this is the amount of information of the input message per bit
of the codeword.
An important notion in coding theory is that of the distance between two codewords;
here (and typically) we use either the Hamming distance ∆(x, y) = #¦i[x
i
= y
i
¦ or the
normalized (Hamming) distance δ(x, y) = ∆(x, y)/n (where n = [x[ = [y[).
The distance d of a code C is the minimum distance between two distinct codewords,
i.e., min
x=y
¦∆(C(x), C(y))¦. For a code of distance d, if the number of bits flipped is
strictly less than d/2 (i.e., weight(η) ≤ d/2), then E(m) +η can be uniquely decoded to m.
A code C : ¦0, 1¦
k
→ ¦0, 1¦
n
with distance d is called an (n, k, d)
2
-code, where n is the size
of the codewords, k the size of the input, d the distance of the code, and 2 the size of the
alphabet ¦0, 1¦.
A code C is called linear if for all x and y, if x and y are codewords then so is x + y
(where addition is performed bitwise modulo 2 – i.e. XOR)
1
. To indicate that a (n, k, d)
2
-
code is also linear, we use the notation [n, k, d]
2
-code (with square brackets). The word
“code,” is commonly used to refer to both the encoding function C (as stated above), as
well as the set of all codewords Im(C).
Typically, the most significant trade-off in coding theory is that between the rate and
distance of a code. For example, given a particular rate, we might ask what is the best
1
Note that if we work over a larger alphabet than binary – e.g. over a larger finite field, we require the
additional constraint that x is a codeword then so are all scalar multiples of x (i.e., αx)
3-1
distance that can be achieved.
3.1.1 Algorithmic Questions and Sublinear Time Algorithms
There are three main algorithmic questions that arise in coding theory:
1. Complexity of encoding;
2. Error detection: given r ∈ ¦0, 1¦
n
, decide if r is a codeword; and
3. Error correction: given r ∈ ¦0, 1¦
n
, find x ∈ ¦0, 1¦
k
minimizing ∆(r, C(x)).
In this course, we will interested in these questions in the context of sublinear time
algorithms. We need to be specific what we mean by sublinear time. Note that a sublinear
time algorithm A to compute a function f ¦0, 1¦
k
→ ¦0, 1¦
n
doesn’t have enough time to
read it’s input or write it’s output. We get over this by accessing the input and writing the
output by oracle access. That is, an oracle machine A is a sublinear time algorithm for f if
A
x
(j) = f(x)
j
where f(x)
j
is the j-th bit of f(x). Note that j need only be log n bits, so
can be read entirely in sublinear time. More formally,
• The input is represented implicitly by an oracle. Whenever the sublinear time algo-
rithm wants to access the j
th
bit of the input string x (for some j ∈ [k]), it queries
the input x-oracle for the j
th
bit and obtains x
j
.
Figure 1: Sublinear Time Algorithms
• The output is not explicitly written by the algorithm, instead it is only implicitly
given by the algorithm. Formally, on being queried for index i of the output string
f(x) (for some i ∈ [n]), the algorithm outputs the bit f(x)
i
. Thus, the algorithm itself
behaves as an oracle for the string f(x), which in turn has oracle access to the input
oracle x.
• Since the algorithm does not read the entire input x, we cannot expect it compute
the output f(x) exactly. We instead relax our guarantee on the output as follows:
On input x ∈ ¦0, 1¦
k
, the algorithm must compute f(x

) exactly for some x

∈ ¦0, 1¦
k
that is -close to the actual input x. In other words, the algorithm computes functions
on some approximation to the input instead of the input itself.
3-2
Property testing, for those familiar with it, is typically done in this framework. Figure 1
gives a pictorial description of a sublinear time algorithm with the above mentioned relax-
ations.
Now we’ll consider the algorithmic questions of coding theory in the sublinear time
framework. Let’s consider the questions above one by one:
1. For a code to have good error-correcting properties, most bits of the codeword needs
to depend on most message-bits. Taking this into account, it does not seem reasonable
to expect a “good” code to have sublinear time encoding. (However, there has been
some recent work in the area of locally encodable codes by relaxing some of the error-
correction properties of the code).
2. Error detection: in this context, error detection is known as local testing. That is,
given r ∈ ¦0, 1¦
n
, test if either r is a codeword or r is far from every codeword (similar
to the gap problems we saw earlier, and also to property testing).
3. Error correction is known as local decoding in this context. Given j, the decoder
queries some bits of a noisy codeword C(x) + η and outputs the j-th bit of x. more
formally, we say a code is locally decodable if there exists a local decoder Dec such
that for all x and r where ∆(r, C(x)) < , the decoder satisfies
∀i ∈ ¦1, . . . , k¦, Pr[Dec
r
(i) = x
i
] ≥
3
4
.
If there is such a decoder for a given code C and it makes fewer than q queries, then
we say C is (q, )-locally decodable.
Obviously, sublinear time algorithms in general, and local decoding and local testing
algorithms in particular, will be randomized algorithms.
Note that local decoding is only required to work when the input is sufficiently close to a
codeword, whereas local testability determines whether a given string is close to a codeword
or far from any codeword. Thus the local decodability of a code says nothing about its local
testability.
3.2 The Walsh-Hadamard Code and Linearity Testing
Now onto our first code, the Walsh-Hadamard code. This will be the main tool in proving
that NP has exponential size PCPs, i.e. NP ⊆ PCP[poly, O(1)]. There are two dual views
of the Walsh-Hadamard code: on the one hand, WH(x) is the evaluation of every linear
function at x; on the other hand, it consists of the dot product of x with every ∈ ¦0, 1¦
n
.
A function f ¦0, 1¦
k
→ ¦0, 1¦ is linear if ∀x, y, f(x + y) = f(x) + f(y).
For example, for any a ∈ ¦0, 1¦
k
, the function
a
(x) =

a
i
x
i
mod 2 (i.e. the dot
product of a and x as vectors over GF(2)) is a linear function. In fact, L
k
= ¦
a
[a ∈ ¦0, 1¦
k
¦
is the set of all linear boolean functions. (Proof: observe that the space of linear functions
is a vector space, and has the same dimension as L
k
.)
The Walsh-Hadamard code WH : ¦0, 1¦
k
→ ¦0, 1¦
2
k
is then defined as WH(x) =
x
,
i.e. WH(x) is the truth table of
x
. More formally, for any a ∈ ¦0, 1¦
k
, the a-th bit of the
Walsh-Hadamard codeword WH(x) is WH(x)
a
=
x
(a).
3-3
Since any two distinct linear functions disagree on exactly half the set of inputs (i.e,
Pr
a
[
x
(a) =
y
(a)] = 1/2, for x = y), the fractional distance of the Walsh-Hadamard codes
is 1/2. Thus the Walsh-Hadamard code is a [k, 2
k
, 2
k−1
]-code. It has very good distance,
but poor rate.
It is useful to note that there are two dual views of the Walsh-Hadamard code based on
the fact that WH(x)
a
=
x
(a) =
a
(x). Thus, WH(x) is both the evaluation of the linear
function
x
at every point as well as the evaluation of every linear function at the point x.
Note that the WH code has the special property that the input bits x
i
are in fact a subset
of the codeword bits, since x
i
=
x
(e
i
). Codes with this property are called systematic codes.
3.2.1 Local Decodability of the Walsh-Hadamard Code
Decoding the Walsh-Hadamard code is very simple. Given a garbled codeword f : ¦0, 1¦
k

¦0, 1¦, which is δ-close to some Walsh-Hadamard codeword, the decoder Dec works as
follows: (The decoder Dec has oracle access to the function f)
Dec
f
: “On input z,
1. Choose r ∈
R
¦0, 1¦
k
2. Query f(z + r) and f(r)
3. Output f(z + r) −f(r)

Claim 3.1. If f : ¦0, 1¦
k
→ ¦0, 1¦ is δ-close to WH(x), then,
Pr[Dec
f
(z) =
x
(z)] ≥ 1 −2δ, ∀x ∈ ¦0, 1¦
k
.
Proof. Since f is δ-close to
x
, we have that for a random r, the probability that f(z +r) =

x
(z + r) is at least 1 − δ, and so is the probability that f(r) =
x
(r). If both of these
conditions hold, then Dec
f
(z) =
x
(z), by the linearity of
x
. Thus Pr
r
[Dec
f
(z) =
x
(z)] ≥
1 −2δ.
Since the decoder only makes two queries, the Walsh-Hadamard code is 2-locally decod-
able.
3.2.2 Local Testability of the Walsh-Hadamard Code: Linearity Testing
To locally test the WH code, we wish to test whether a given truth table is the truth table
of a linear function. The problem of local testing of the WH code is more commonly called
linearity testing. Formally, given f ¦0, 1¦
k
→ ¦0, 1¦ we want to test whether it is a linear
function (WH codeword) or far from linear.
The test is, as in the WH decoder, quite simple: pick y and z uniformly at random from
¦0, 1¦
k
and check that f(z) + f(y) = f(z + y). This test was proposed and first analyzed
by Blum, Luby and Rubinfeld [BLR93].
BLR-Test
f
: “ 1. Choose y, z ∈
R
¦0, 1¦
k
independently
2. Query f(y), f(z), and f(y + z)
3. Accept if f(y) + f(z) = f(y + z).

3-4
Obviously if f is linear, this test will pass with probability 1. The question is, can there
be function that are far from linear but still pass this test with high probability? No, as
shown in the following
Theorem 3.2. If f is δ-far from linear, then
Pr[BLR-Test
f
accepts ] ≤ 1 −δ.
The above theorem is tight since a random function is 1/2-far from linear, and passes
the BLR-Test with probability exactly 1/2.
The original proof of Blum, Luby and Rubinfeld [BLR93] is a combinatorial proof of
a weaker version of the above theorem, but we will give an algebraic proof, as similar
techniques will arise later in the course. This algebraic proof is due to Bellare, Coppersmith,
h˚astad, Kiwi and Sudan [BCH
+
96]. Before proceeding to the proof, we will first need to
equip ourselves with some basics of Boolean Fourier analysis.
3.2.3 Fourier Analysis
First, rather than working over ¦0, 1¦ as the output of a linear function, it will be convenient
to treat the output space as ¦+1, −1¦ (the roots of unity) under multiplication. Thus the
Boolean 0 corresponds to 1 in this setting, and the Boolean 1 corresponds to -1. Linearity
now takes the form: f ¦0, 1¦
n
→ ¦+1, −1¦ is linear if f(x + y) = f(x)f(y).
Consider the family of functions T = ¦f : ¦0, 1¦
k
→ R¦ and equip T with an addition
(f + g)(x) = f(x) + g(x). It is clear that T is a vector space over the reals. Furthermore,
the characteristic functions ¦δ
a
[a ∈ ¦0, 1¦
k
¦ are a basis, where δ
a
(x) = 1 if x = a and 0
otherwise. Thus T has dimension 2
k
.
We will now show that the linear functions of the form χ
a
(x) = (−1)
a(x)
= (−1)
P
a
i
x
i
( mod 2)
also form a basis for T. For this, it is fruitful to define an inner product on T as follows:
'f, g` = Exp
x∈{0,1}
k [f(x)g(x)] =
1
2
k

x∈{0,1}
k
f(x)g(x).
(It is an easy exercise to check that this is in fact an inner product.) Note that there are
already 2
k
= dimT functions of this form, so all we need to do is show that they are linearly
independent.
We begin by examining a few basic properties of the functions χ
a
. First, note that
Property 1. χ
a
(x + y) = χ
a
(x)χ
a
(y)
i.e. χ
a
is linear, as mentioned previously. Second,
Property 2. χ
a+b
(x) = χ
a−b
(x) = χ
a
(x)χ
b
(x)
i.e. χ
a
(x) is also linear in a; this should come as no surprise, because of the duality of

a
(x) as both a linear function of a and of x.
The first property we will need that is not entirely obvious is that
Property 3.
Exp
x

a
(x)] =

1 if a = 0
0 otherwise
3-5
Proof. If a = 0, this clearly holds. If a = 0, then by permuting the indices we may assume
that a
1
= 0 without loss of generality. Then we have
2
k
Exp
x

a
(x)] =

x
(−1)
P
a
i
x
i
=

x:x
1
=1
(−1)
P
a
i
x
i
+

x:x
1
=0
(−1)
P
a
i
x
i
Then, since (−1)
a(0y)
= −(−1)
a(1y)
, these two sums exactly cancel out.
We then have,
Property 4.

a
, χ
b
` =

1 if a = b
0 otherwise
.
This follows from the above via 'χ
a
, χ
b
` = E
x

a
(x)χ
b
(x)] = E
x

a−b
(x)], and then
applying the above fact.
Thus, the χ
a
form not only a basis for T, but also an orthonormal basis for T. Since
the χ
a
form an orthonormal basis, for any f ∈ T we may write f =

a
ˆ
f
a
χ
a
for 'f, χ
a
` =
ˆ
f
a
∈ R. These
ˆ
f
a
are called the Fourier coefficients of f.
Observer that if the normalized distance δ(f, χ
a
) = , then
ˆ
f
a
= 1(1−)+(−1) = 1−2,
so the Fourier coefficients capture the normalized distance from linear functions.
Now we come to one of the most basic useful facts in Fourier analysis:
Property 5 (Parseval’s identity). 'f, f` =

ˆ
f
2
a
Proof. Writing f in terms of the basis χ
a
, we get:
'f, f` = '

a
ˆ
f
a
χ
a
,

b
ˆ
f
b
χ
b
`
=

a,b
ˆ
f
a
ˆ
f
b

a
, χ
b
` (by linearity of ', `)
=

a
ˆ
f
2
a
where the last line follows from the previous because the χ
a
form an orthonormal basis.
Corollary 3.3. In particular, if f is a Boolean function, i.e. ±1-valued, then

ˆ
f
2
a
= 1.
3.2.4 Proof of Soundness of BLR-Test
Finally, we come to the proof of the soundness of the Blum-Luby-Rubinfeld linearity test:
Proof of Theorem 3.2. Suppose f is δ-far from any linear function. Note that we can rewrite
the linearity condition f(x)f(y) = f(x + y) as f(x)f(y)f(x + y) = 1, since f is ±1-valued.
Then
Pr
x,y
[BLR-Test accepts f] = Pr
x,y
[f(x)f(y)f(x + y) = 1]
3-6
Note that for any random variable Z with values in ¦+1, −1¦, Pr(Z = 1) = Exp[
1+Z
2
],
since if Z = 1, then (1 + Z)/2 = 1 and if Z = −1, then (1 + Z)/2 = 0, so (1 + Z)/2 is an
indicator variable for the event Z = 1. Thus we have
Pr
x,y
[BLR-Test accepts f] = Exp
x,y

1 + f(x)f(y)f(x + y)
2

=
1
2
+
1
2
Exp
x,y
[f(x)f(y)f(x + y)]
Now, writing out f in terms of its Fourier coefficients, we get
Pr
x,y
[BLR-Test accepts f] =
1
2
+
1
2
Exp
x,y

a,b,c
ˆ
f
a
ˆ
f
b
ˆ
f
c
χ
a
(x)χ
b
(y)χ
c
(x + y)


=
1
2
+
1
2
Exp
x,y

a,b,c
ˆ
f
a
ˆ
f
b
ˆ
f
c
χ
a
(x)χ
b
(y)χ
c
(x)χ
c
(y)


Then, we apply linearity of expectation and the fact that x and y are independent to get:
Pr
x,y
[BLR accepts f] =
1
2
+
1
2

a,b,c
ˆ
f
a
ˆ
f
b
ˆ
f
c
Exp
x

a
(x)χ
c
(x)] Exp
y

b
(y)χ
c
(y)]
=
1
2
+
1
2

a,b,c
ˆ
f
3
a

1
2
+
1
2
max
a
(
ˆ
f
a
)

a
ˆ
f
2
a
=
1
2
+
1
2
max
a
(
ˆ
f
a
) (by Parseval’s identity)
= 1 −δ
where the last line follows from the fact that f is δ-far from linear, so its largest Fourier
coefficient can be at most 1 −2δ, as noted previously.
References
[BCH
+
96] Mihir Bellare, Don Coppersmith, Johan H˚astad, Marcos A. Kiwi, and
Madhu Sudan. Linearity testing in characteristic two. IEEE Transactions on
Information Theory, 42(6):1781–1795, November 1996. (Preliminary version in
36th FOCS, 1995). doi:10.1109/18.556674.
[BLR93] Manuel Blum, Michael Luby, and Ronitt Rubinfeld. Self-
testing/correcting with applications to numerical problems. J. Computer and
System Sciences, 47(3):549–595, December 1993. (Preliminary Version in 22nd
STOC, 1990). doi:10.1016/0022-0000(93)90044-W.
3-7

1}k → {0. C(x)). an oracle machine A is a sublinear time algorithm for f if Ax (j) = f (x)j where f (x)j is the j-th bit of f (x). Whenever the sublinear time algorithm wants to access the j th bit of the input string x (for some j ∈ [k]). We instead relax our guarantee on the output as follows: On input x ∈ {0. the algorithm outputs the bit f (x)i .1. so can be read entirely in sublinear time. • The input is represented implicitly by an oracle. We need to be specific what we mean by sublinear time. Formally.1 Algorithmic Questions and Sublinear Time Algorithms There are three main algorithmic questions that arise in coding theory: 1. which in turn has oracle access to the input oracle x. find x ∈ {0. we cannot expect it compute the output f (x) exactly. 1}n . the algorithm must compute f (x ) exactly for some x ∈ {0. the algorithm itself behaves as an oracle for the string f (x). 1}k that is -close to the actual input x.distance that can be achieved. More formally. and 3. That is. We get over this by accessing the input and writing the output by oracle access. • Since the algorithm does not read the entire input x. on being queried for index i of the output string f (x) (for some i ∈ [n]). the algorithm computes functions on some approximation to the input instead of the input itself. 2. In this course. 3-2 . In other words. Complexity of encoding. Error detection: given r ∈ {0. Error correction: given r ∈ {0. 1}n doesn’t have enough time to read it’s input or write it’s output. 1}k . Note that j need only be log n bits. instead it is only implicitly given by the algorithm. Note that a sublinear time algorithm A to compute a function f {0. 1}n . we will interested in these questions in the context of sublinear time algorithms. 1}k minimizing ∆(r. Thus. 3. decide if r is a codeword. it queries the input x-oracle for the j th bit and obtains xj . Figure 1: Sublinear Time Algorithms • The output is not explicitly written by the algorithm.

the a-th bit of the Walsh-Hadamard codeword W H(x) is W H(x)a = x (a). the decoder satisfies 3 ∀i ∈ {1. Obviously. O(1)]. k}. W H(x) is the truth table of x . More formally. 1}k → {0. Given j. most bits of the codeword needs to depend on most message-bits. and also to property testing). Error correction is known as local decoding in this context. 3. sublinear time algorithms in general. 1}n . 4 If there is such a decoder for a given code C and it makes fewer than q queries. This will be the main tool in proving that N P has exponential size PCPs. A function f {0. error detection is known as local testing. . Thus the local decodability of a code says nothing about its local testability.2 The Walsh-Hadamard Code and Linearity Testing Now onto our first code. There are two dual views of the Walsh-Hadamard code: on the one hand. )-locally decodable. y. for any a ∈ {0. . 1}k } is the set of all linear boolean functions.) k The Walsh-Hadamard code W H : {0. 3. the dot product of a and x as vectors over GF (2)) is a linear function. and local decoding and local testing algorithms in particular.e. C(x)) < . In fact. we say a code is locally decodable if there exists a local decoder Dec such that for all x and r where ∆(r. 2. Taking this into account. for those familiar with it. 1}2 is then defined as W H(x) = x . Figure 1 gives a pictorial description of a sublinear time algorithm with the above mentioned relaxations. . W H(x) is the evaluation of every linear function at x. given r ∈ {0. and has the same dimension as Lk . 1} is linear if ∀x. Lk = { a |a ∈ {0. i. 1}n . 1}k → {0. it consists of the dot product of x with every ∈ {0.e. f (x + y) = f (x) + f (y). Error detection: in this context. test if either r is a codeword or r is far from every codeword (similar to the gap problems we saw earlier. Let’s consider the questions above one by one: 1. i. there has been some recent work in the area of locally encodable codes by relaxing some of the errorcorrection properties of the code). 1}k . That is. on the other hand. 1}k . the decoder queries some bits of a noisy codeword C(x) + η and outputs the j-th bit of x. For a code to have good error-correcting properties.e. .Property testing. more formally. Now we’ll consider the algorithmic questions of coding theory in the sublinear time framework. Note that local decoding is only required to work when the input is sufficiently close to a codeword. (Proof: observe that the space of linear functions is a vector space. will be randomized algorithms. the function a (x) = ai xi mod 2 (i. (However. the Walsh-Hadamard code. is typically done in this framework. for any a ∈ {0. it does not seem reasonable to expect a “good” code to have sublinear time encoding. Pr[Decr (i) = xi ] ≥ . whereas local testability determines whether a given string is close to a codeword or far from any codeword. then we say C is (q. For example. 3-3 . N P ⊆ P CP [poly.

It is useful to note that there are two dual views of the Walsh-Hadamard code based on the fact that W H(x)a = x (a) = a (x). If f : {0. we have that for a random r. we wish to test whether a given truth table is the truth table of a linear function. 1}k → {0.2.1 Local Decodability of the Walsh-Hadamard Code Decoding the Walsh-Hadamard code is very simple. f (z). Query f (z + r) and f (r) 3. 1} is δ-close to W H(x). by the linearity of x . the probability that f (z + r) = x (z + r) is at least 1 − δ. Since f is δ-close to x .1. the fractional distance of the Walsh-Hadamard codes is 1/2.2. quite simple: pick y and z uniformly at random from {0. Thus Prr [Decf (z) = x (z)] ≥ 1 − 2δ. Accept if f (y) + f (z) = f (y + z). Choose y. ∀x ∈ {0. 2k−1 ]-code. Since the decoder only makes two queries. 1}k and check that f (z) + f (y) = f (z + y). 3-4 . The test is. Thus the Walsh-Hadamard code is a [k. Given a garbled codeword f : {0. 3. the Walsh-Hadamard code is 2-locally decodable. the decoder Dec works as follows: (The decoder Dec has oracle access to the function f ) Decf : “On input z. 1}k 2. z ∈R {0.2 Local Testability of the Walsh-Hadamard Code: Linearity Testing To locally test the WH code. and f (y + z) 3. since xi = x (ei ). Choose r ∈R {0. 1}k independently 2. Luby and Rubinfeld [BLR93]. 1. Thus. Formally. which is δ-close to some Walsh-Hadamard codeword. Codes with this property are called systematic codes. If both of these conditions hold. 2k . 3. then. BLR-Testf : “ 1. given f {0. 1}k → {0. Proof. Output f (z + r) − f (r) Claim 3. 1}. This test was proposed and first analyzed by Blum.Since any two distinct linear functions disagree on exactly half the set of inputs (i. Note that the WH code has the special property that the input bits xi are in fact a subset of the codeword bits. and so is the probability that f (r) = x (r). 1} we want to test whether it is a linear function (WH codeword) or far from linear. then Decf (z) = x (z). The problem of local testing of the WH code is more commonly called linearity testing.e. Pra [ x (a) = y (a)] = 1/2. It has very good distance. W H(x) is both the evaluation of the linear function x at every point as well as the evaluation of every linear function at the point x. Query f (y). 1}k → {0. for x = y). but poor rate. as in the WH decoder. Pr[Decf (z) = x (z)] ≥ 1 − 2δ. 1}k .

1} as the output of a linear function. as similar techniques will arise later in the course. Before proceeding to the proof. where δa (x) = 1 if x = a and 0 otherwise. Kiwi and Sudan [BCH+ 96]. it is fruitful to define an inner product on F as follows: f. then Pr[BLR-Testf accepts ] ≤ 1 − δ. First. x∈{0.) Note that there are already 2k = dim F functions of this form.1}k [f (x)g(x)] = 1 · 2k f (x)g(x). Thus the Boolean 0 corresponds to 1 in this setting. we will first need to equip ourselves with some basics of Boolean Fourier analysis. the characteristic functions {δa |a ∈ {0. Consider the family of functions F = {f : {0. 1}k → R} and equip F with an addition (f + g)(x) = f (x) + g(x). a The first property we will need that is not entirely obvious is that Property 3. It is clear that F is a vector space over the reals. Thus F has dimension 2k . −1} is linear if f (x + y) = f (x)f (y). this should come as no surprise. The original proof of Blum. h˚ astad. as shown in the following Theorem 3. χa+b (x) = χa−b (x) = χa (x)χb (x) i. χa (x + y) = χa (x)χa (y) i. but we will give an algebraic proof. Linearity now takes the form: f {0. note that Property 1. 3. can there be function that are far from linear but still pass this test with high probability? No.e. This algebraic proof is due to Bellare. The above theorem is tight since a random function is 1/2-far from linear. Furthermore. Coppersmith.1}k (It is an easy exercise to check that this is in fact an inner product. it will be convenient to treat the output space as {+1. −1} (the roots of unity) under multiplication. as mentioned previously. P We will now show that the linear functions of the form χa (x) = (−1) a (x) = (−1) ai xi ( mod 2) also form a basis for F. Second. Expx [χa (x)] = 1 if a = 0 0 otherwise 3-5 . We begin by examining a few basic properties of the functions χa . Luby and Rubinfeld [BLR93] is a combinatorial proof of a weaker version of the above theorem. this test will pass with probability 1. The question is.2. because of the duality of (x) as both a linear function of a and of x. Property 2. so all we need to do is show that they are linearly independent. If f is δ-far from linear. 1}k } are a basis. 1}n → {+1. and passes the BLR-Test with probability exactly 1/2. rather than working over {0. and the Boolean 1 corresponds to -1.2. For this.e. χa (x) is also linear in a.Obviously if f is linear. g = Expx∈{0.3 Fourier Analysis First. χa is linear.

Writing f in terms of the basis χa . b ˆ fb χb = a. ±1-valued. so the Fourier coefficients capture the normalized distance from linear functions. If a = 0. we get: f. since (−1) a (0y) (−1) x:x1 =1 P ai xi + (−1) x:x1 =0 P ai xi = −(−1) a (1y) . then by permuting the indices we may assume that a1 = 0 without loss of generality. Since ˆ the χa form an orthonormal basis. Now we come to one of the most basic useful facts in Fourier analysis: Property 5 (Parseval’s identity). for any f ∈ F we may write f = a fa χa for f.2. then ˆ2 fa = 1. but also an orthonormal basis for F. since f is ±1-valued. f = a ˆ fa χa . these two sums exactly cancel out. Suppose f is δ-far from any linear function. Note that we can rewrite the linearity condition f (x)f (y) = f (x + y) as f (x)f (y)f (x + y) = 1. 3. and then applying the above fact. χb = Ex [χa (x)χb (x)] = Ex [χa−b (x)]. if f is a Boolean function. ˆ Observer that if the normalized distance δ(f. χa ) = . These fa are called the Fourier coefficients of f . Then Pr [BLR-Test accepts f ] = Pr [f (x)f (y)f (x + y) = 1] x.2.4 Proof of Soundness of BLR-Test Finally. i. this clearly holds. χb (by linearity of ·. In particular. χa . Thus. Property 4. f.Proof. Corollary 3. then fa = 1(1− )+(−1) = 1−2 . · ) ˆ2 fa a = where the last line follows from the previous because the χa form an orthonormal basis.b ˆ ˆ fa fb χa . If a = 0. χa = ˆ ˆ fa ∈ R. we come to the proof of the soundness of the Blum-Luby-Rubinfeld linearity test: Proof of Theorem 3. We then have. otherwise This follows from the above via χa .e. χb = 1 0 if a = b . f = ˆ2 fa Proof.y 3-6 . Then we have 2k · Expx [χa (x)] = x (−1) P ai xi = Then.3. the χa form not only a basis for F.y x.

Johan H˚ astad. P r(Z = 1) = Exp[ 1+Z ]. and Madhu Sudan. IEEE Transactions on Information Theory. Michael Luby.b. we get   1 1 ˆ ˆˆ Pr [BLR-Test accepts f ] = + Expx.y x. J.1109/18.y 1 + f (x)f (y)f (x + y) 2 = 1 1 + Expx. 47(3):549–595. then (1 + Z)/2 = 1 and if Z = −1. 1995).556674. doi:10.y 1 1 + 2 2 1 1 + 2 2 ˆ ˆˆ fa fb fc Expx [χa (x)χc (x)] Expy [χb (y)χc (y)] a. Don Coppersmith. References [BCH+ 96] Mihir Bellare. so its largest Fourier coefficient can be at most 1 − 2δ.b. doi:10. as noted previously. 3-7 . November 1996.c Then. December 1993.b. [BLR93] Manuel Blum. writing out f in terms of its Fourier coefficients.y  fa fb fc χa (x)χb (y)χc (x + y) x. we apply linearity of expectation and the fact that x and y are independent to get: Pr [BLR accepts f ] = x. Thus we have Pr [BLR-Test accepts f ] = Expx.b. 42(6):1781–1795. and Ronitt Rubinfeld. Kiwi.c 1 1 ˆ + max(fa ) 2 2 a ˆ2 fa a 1 1 ˆ + max(fa ) (by Parseval’s identity) = 2 2 a = 1−δ where the last line follows from the fact that f is δ-far from linear. (Preliminary Version in 22nd STOC. Computer and System Sciences. so (1 + Z)/2 is an indicator variable for the event Z = 1. 1990).y 2 2 a. Linearity testing in characteristic two. (Preliminary version in 36th FOCS. 2 since if Z = 1. −1}.Note that for any random variable Z with values in {+1.1016/0022-0000(93)90044-W. Selftesting/correcting with applications to numerical problems. then (1 + Z)/2 = 0. Marcos A.y  fa fb fc χa (x)χb (y)χc (x)χc (y) 2 2 a.y [f (x)f (y)f (x + y)] 2 2 Now.c = ≤ ˆ3 fa a.c   1 1 ˆ ˆˆ = + Expx.