Small-bias sample space
In theoretical computer science, a small-bias sample space (also known as -biased sample space,
-biased generator, or small-bias probability space) is a probability distribution that fools parity functions.
In other words, no parity function can distinguish between a small-bias sample space and the uniform distribution with high probability, and hence, small-bias sample spaces naturally give rise to pseudorandom generators for parity functions.
The main useful property of small-bias sample spaces is that they need far fewer truly random bits than the uniform distribution to fool parities. Efficient constructions of small-bias sample spaces have found many applications in computer science, some of which are derandomization, error-correcting codes, and probabilistically checkable proofs.
The connection with error-correcting codes is in fact very strong since -biased sample spaces are equivalent to
-balanced error-correcting codes.
Definition
Bias
Let be a probability distribution over
.
The bias of
with respect to a set of indices
is defined as[1]
where the sum is taken over , the finite field with two elements. In other words, the sum
equals
if the number of ones in the sample
at the positions defined by
is even, and otherwise, the sum equals
.
For
, the empty sum is defined to be zero, and hence
.
ϵ-biased sample space
A probability distribution over
is called an
-biased sample space if
holds for all non-empty subsets
.
ϵ-biased set
An -biased sample space
that is generated by picking a uniform element from a multiset
is called
-biased set.
The size
of an
-biased set
is the size of the multiset that generates the sample space.
ϵ-biased generator
An -biased generator
is a function that maps strings of length
to strings of length
such that the multiset
is an
-biased set. The seed length of the generator is the number
and is related to the size of the
-biased set
via the equation
.
Connection with epsilon-balanced error-correcting codes
There is a close connection between -biased sets and
-balanced linear error-correcting codes.
A linear code
of message length
and block length
is
-balanced if the Hamming weight of every nonzero codeword
is between
and
.
Since
is a linear code, its generator matrix is an
-matrix
over
with
.
Then it holds that a multiset is
-biased if and only if the linear code
, whose columns are exactly elements of
, is
-balanced.[2]
Constructions of small epsilon-biased sets
Usually the goal is to find -biased sets that have a small size
relative to the parameters
and
.
This is because a smaller size
means that the amount of randomness needed to pick a random element from the set is smaller, and so the set can be used to fool parities using few random bits.
Theoretical bounds
The probabilistic method gives a non-explicit construction that achieves size .[2]
The construction is non-explicit in the sense that finding the
-biased set requires a lot of true randomness, which does not help towards the goal of reducing the overall randomness.
However, this non-explicit construction is useful because it shows that these efficient codes exist.
On the other hand, the best known lower bound for the size of
-biased sets is
, that is, in order for a set to be
-biased, it must be at least that big.[2]
Explicit constructions
There are many explicit, i.e., deterministic constructions of -biased sets with various parameter settings:
- Naor & Naor (1990) achieve
. The construction makes use of Justesen codes (which is a concatenation of Reed–Solomon codes with the Wozencraft ensemble) as well as expander walk sampling.
- Alon et al. (1992) achieve
. One of their constructions is the concatenation of Reed–Solomon codes with the Hadamard code; this concatenation turns out to be an
-balanced code, which gives rise to an
-biased sample space via the connection mentioned above.
- Concatenating Algebraic geometric codes with the Hadamard code gives an
-balanced code with
.[2]
- Ben-Aroya & Ta-Shma (2009) achieves
.
These bounds are mutually incomparable. In particular, none of these constructions yields the smallest -biased sets for all settings of
and
.
Application: almost k-wise independence
An important application of small-bias sets lies in the construction of almost k-wise independent sample spaces.
k-wise independent spaces
A random variable over
is a k-wise independent space if, for all index sets
of size
, the marginal distribution
is exactly equal to the uniform distribution over
.
That is, for all such
and all strings
, the distribution
satisfies
.
Constructions and bounds
k-wise independent spaces are fairly well-understood.
- A simple construction by Joffe (1974) achieves size
.
- Alon, Babai & Itai (1986) construct a k-wise independent space whose size is
.
- Chor et al. (1985) prove that no k-wise independent space can be significantly smaller than
.
Joffe's construction
Joffe (1974) constructs a -wise independent space
over the finite field with some prime number
of elements, i.e.,
is a distribution over
. The initial
marginals of the distribution are drawn independently and uniformly at random:
.
For each with
, the marginal distribution of
is then defined as
where the calculation is done in .
Joffe (1974) proves that the distribution
constructed in this way is
-wise independent as a distribution over
.
The distribution
is uniform on its support, and hence, the support of
forms a
-wise independent set.
It contains all
strings in
that have been extended to strings of length
using the deterministic rule above.
Almost k-wise independent spaces
A random variable over
is a
-almost k-wise independent space if, for all index sets
of size
, the restricted distribution
and the uniform distribution
on
are
-close in 1-norm, i.e.,
.
Constructions
Naor & Naor (1990) give a general framework for combining small k-wise independent spaces with small -biased spaces to obtain
-almost k-wise independent spaces of even smaller size.
In particular, let
be a linear mapping that generates a k-wise independent space and let
be a generator of an
-biased set over
.
That is, when given a uniformly random input, the output of
is a k-wise independent space, and the output of
is
-biased.
Then
with
is a generator of an
-almost
-wise independent space, where
.[3]
As mentioned above, Alon, Babai & Itai (1986) construct a generator with
, and Naor & Naor (1990) construct a generator
with
.
Hence, the concatenation
of
and
has seed length
.
In order for
to yield a
-almost k-wise independent space, we need to set
, which leads to a seed length of
and a sample space of total size
.
Notes
- ↑ cf., e.g., Goldreich (2001)
- 1 2 3 4 cf., e.g., p. 2 of Ben-Aroya & Ta-Shma (2009)
- ↑ Section 4 in Naor & Naor (1990)
References
- Alon, Noga; Babai, László; Itai, Alon (1986), "A fast and simple randomized parallel algorithm for the maximal independent set problem" (PDF), Journal of Algorithms 7 (4): 567–583, doi:10.1016/0196-6774(86)90019-2
- Alon, Noga; Goldreich, Oded; Håstad, Johan; Peralta, René (1992), "Simple Constructions of Almost k-wise Independent Random Variables" (PDF), Random Structures & Algorithms 3 (3): 289–304, doi:10.1002/rsa.3240030308
- Ben-Aroya, Avraham; Ta-Shma, Amnon (2009), "Constructing Small-Bias Sets from Algebraic-Geometric Codes" (PDF), Proceedings of the 50th Annual Symposium on Foundations of Computer Science, FOCS 2009: 191–197, doi:10.1109/FOCS.2009.44, ISBN 978-1-4244-5116-6
- Chor, Benny; Goldreich, Oded; Håstad, Johan; Freidmann, Joel; Rudich, Steven; Smolensky, Roman (1985), "The bit extraction problem or t-resilient functions", Proceedings of the 26th Annual Symposium on Foundations of Computer Science, FOCS 1985: 396–407, doi:10.1109/SFCS.1985.55, ISBN 0-8186-0644-4
- Goldreich, Oded (2001), Lecture 7: Small bias sample spaces
- Joffe, Anatole (1974), "On a Set of Almost Deterministic k-Independent Random Variables", Annals of Probability 2 (1): 161–162, doi:10.1214/aop/1176996762
- Naor, Joseph; Naor, Moni (1990), "Small-bias Probability Spaces: efficient constructions and Applications", Proceedings of the 22nd annual ACM symposium on Theory of computing, STOC 1990: 213–223, doi:10.1145/100216.100244, ISBN 0897913612