Decoding Downset codes over a finite grid

by   Srikanth Srinivasan, et al.
IIT Bombay

In a recent paper, Kim and Kopparty (Theory of Computing, 2017) gave a deterministic algorithm for the unique decoding problem for polynomials of bounded total degree over a general grid. We show that their algorithm can be adapted to solve the unique decoding problem for the general family of Downset codes. Here, a downset code is specified by a family D of monomials closed under taking factors: the corresponding code is the space of evaluations of all polynomials that can be written as linear combinations of monomials from D.



page 1

page 2

page 3

page 4


On decoding procedures of intertwining codes

One of the main weakness of the family of centralizer codes is that its ...

A Decoding Algorithm for Rank Metric Codes

In this work we will present algorithms for decoding rank metric codes. ...

Beyond the Guruswami-Sudan (and Parvaresh-Vardy) Radii: Folded Reed-Solomon, Multiplicity and Derivative Codes

The classical family of Reed-Solomon codes consist of evaluations of pol...

Erasure decoding of convolutional codes using first order representations

In this paper, we employ the linear systems representation of a convolut...

On decoding hyperbolic codes

Few decoding algorithms for hyperbolic codes are known in the literature...

Orbit Structure of Grassmannian G_2, m and a decoder for Grassmann code C(2, m)

In this manuscript, we consider decoding Grassmann codes, linear codes a...

A New Path to Code-based Signatures via Identification Schemes with Restricted Errors

In this paper we introduce a variant of the Syndrome Decoding Problem (S...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

1 Preliminaries

Throughout fix a field Let be finite non-empty subsets of and let denote the grid We use to denote Given functions , we use to denote the Hamming distance between and , i.e. the number of points where they differ.

Let denote the set with the natural partial order For each we denote by the set and by the set For any , we will identify the monomial with and use the monomial notation and the multi-index notation interchangeably.

The following fact is standard.

Fact 1.

Each has a unique representation as a polynomial where the degree of in is at most for each . Equivalently, there is a natural one-one correspondence between the space of all functions from to and

Given a downset we associate with it the linear code , called a downset code, defined by

The following lemma allows us to compute the minimum distance for any downset Recall (see e.g. [CLO07]) that a monomial order on monomials in is a total ordering of the monomials that is a well order and moreover satisfies the following for any :

Lemma 2 (Schwartz-Zippel Lemma for ).
  1. Let be arbitrary and let be the leading monomial of w.r.t. a monomial order. Then,

  2. For each there is an such that and can be represented by a linear combination of monomials from In particular, if then such an

Thus, In particular, given it can be found in polynomial time.


Item 1 is an easy consequence of the proof of Macaulay’s theorem [Mac27] (see also [CLO07, Chapter 5.3, Proposition 4]). For completeness, we present a short proof here. Given any polynomial , let denote the set of monomials with non-zero coefficient in . For any set of monomials let denote the set of monomials that divide some monomial in

For any , let Note that is a univariate polynomial of degree that vanishes on . Given any polynomial , the remainder obtained upon dividing by has degree in the variable and evaluates to the same value as at points in . Further, each monomial in divides some monomial in (i.e. ) . Repeating this process, we eventually obtain a polynomial representing the same function as .

Let be the subset of points in where To prove item 1 of the lemma, it suffices to show that every can be represented as a polynomial from Standard linear algebra then implies that which proves item 1.

To prove the above, fix any . By extending in an arbitrary way to , we know that can be represented by some polynomial If does not contain any monomial from then we are done. Otherwise, we choose the largest (w.r.t. ) monomial . Let be the coefficient of in .

Assume that where . Multiplying by , we see that the polynomial vanishes on . Note that is a polynomial such that that also represents the function at points in . Repeating this process, we eventually obtain a polynomial without any monomials from that represents . The polynomial (obtained by dividing by as mentioned above) also represents and furthermore is an element of

For item 2, assume that for each and consider

Let denote with its natural partial order. Let denote the set

Given a downset as above, let For define

The following observation will be useful.

Lemma 3.

Let be any downset, and let

  1. For each is a downset in Further, we have

  2. For each we have

  1. Clear from the definition, since is a downset.

  2. Let and have weight equal to , and and have weight equal to . Let be any monomial in and be any monomial in . So and . Then by Item 1, and so , that is, . Thus, and hence, has weight at least . So we see that

(Note that the quantity is the distance of the degree- Reed-Solomon code on the set )

As in the result of Kim and Kopparty, we work with the more general problem of decoding Weighted functions (or weighted received word). A weighted function over is a function or equivalently a pair where and Given a weighted function and a function , we define their distance by

A (unweighted) function is identified with the weighted function where is the identically zero function. Note that with this identification, agrees with the standard Hamming distance between and . In particular, the unique decoding problem for immediately reduces to the problem of finding a codeword of distance less than from a given weighted function.

We will also need the following lemma about weighted codewords of a downset code . The proof (in a more general setting) can be found in [KK17, Lemma 2.1].

Lemma 4.

Assume that are distinct.333In the algorithm, we will only need this for , i.e. for the Reed-Solomon code. Let be any weighted received word. Then, In particular, both and cannot be at distance strictly less than from .

2 The main theorem

1:Input: , where
  • is a finite grid in with . We have .

  • is a downset.

  • is a weighted received word.

2:if  then
3:     return WeightedRSDecoder(). Here and , for some .
5:     Define and by
6:     for  do
7:         Define by
8:         for  do
9:              Define by , for all .
10:              Let .
11:              if  then
12:                  .
13:                  .
14:              else
15:                  .
16:                  .
17:              end if
18:         end for
19:         Define weighted word and by
20:         Let .
21:     end for
22:     return .
23:end if
Algorithm 1 WeightedDownsetDecoder: Decoding a downset code over a grid
Theorem 5.

There is a deterministic polynomial time algorithm that given, and a weighted codeword produces a codeword such that if one exists. (If no such codeword exists, the algorithm outputs some arbitrary polynomial.)

For the case when , a result of Forney [Jr.66] yields a deterministic polynomial-time algorithm for this problem. We call this algorithm WeightedRSDecoder and refer the reader to [Jr.66] or [KK17] for a description.


The algorithm is specified as WeightedDownsetDecoder below (Algorithm 1). We prove its correctness by induction on .

For the base case , we simply use the algorithm WeightedRSDecoder and so there is nothing to prove.

Now we assume the correctness of WeightedDownsetDecoder algorithm for indeterminates.

Let be a received weighted word. Suppose there is a with . We can write

where as in Algorithm 1.

We show, by induction on , that the algorithm correctly decodes the polynomial In other words, for each the polynomial computed by Algorithm 1 is the same as the polynomial

Fix . Assume that the algorithm has correctly decoded for each . Let . Note that , for all .

To show that it is enough to show that where is as computed by the algorithm. Then, the induction hypothesis implies that .

Define as in the algorithm by

By induction, we know that for . Hence, we observe that

Hence, .

We now analyze Recall that we have

Fix some . Define Also, define

We claim that


We prove (1) by a case analysis.

  1. This implies that In particular, this implies that since otherwise the Reed-Solomon decoder would have returned instead of This immediately implies (1).

    So from now we will assume that By Lemma 4, it follows that

  2. and .

    In this case, we immediately have

  3. and .

    As in the previous case, we have . But as we have . Thus, by Lemma 4, it follows that . Hence

This concludes the proof of (1). Using (1), we get

This completes the proof. ∎


The authors are grateful to Swastik Kopparty and Madhu Sudan for their helpful comments and encouragement.


  • [ALM98] Sanjeev Arora, Carsten Lund, Rajeev Motwani, Madhu Sudan, and Mario Szegedy. Proof verification and the hardness of approximation problems. J. ACM, 45(3):501–555, 1998.
  • [BFL91] László Babai, Lance Fortnow, and Carsten Lund. Non-deterministic exponential time has two-prover interactive protocols. Computational Complexity, 1:3–40, 1991.
  • [BFNW93] László Babai, Lance Fortnow, Noam Nisan, and Avi Wigderson. BPP has subexponential time simulations unless EXPTIME has publishable proofs. Computational Complexity, 3:307–318, 1993.
  • [CLO07] David Cox, John Little, and Donal O’Shea. Ideals, varieties, and algorithms. Undergraduate Texts in Mathematics. Springer, New York, third edition, 2007. An introduction to computational algebraic geometry and commutative algebra.
  • [DL78] Richard A. DeMillo and Richard J. Lipton. A probabilistic remark on algebraic program testing. Inf. Process. Lett., 7(4):193–195, 1978.
  • [Jr.66] G. David Forney Jr. Generalized minimum distance decoding. IEEE Trans. Information Theory, 12(2):125–131, 1966.
  • [KK17] John Y. Kim and Swastik Kopparty. Decoding reed-muller codes over product sets. Theory of Computing, 13(21):1–38, 2017.
  • [LFKN92] Carsten Lund, Lance Fortnow, Howard J. Karloff, and Noam Nisan. Algebraic methods for interactive proof systems. J. ACM, 39(4):859–868, 1992.
  • [Mac27] F. S. Macaulay. Some properties of enumeration in the theory of modular systems. Proceedings of the London Mathematical Society, s2-26(1):531–555, 1927.
  • [Ree54] Irving S. Reed. A class of multiple-error-correcting codes and the decoding scheme. Trans. of the IRE Professional Group on Information Theory (TIT), 4:38–49, 1954.
  • [Sch80] Jacob T. Schwartz. Fast probabilistic algorithms for verification of polynomial identities. J. ACM, 27(4):701–717, 1980.
  • [Sha92] Adi Shamir. IP = PSPACE. J. ACM, 39(4):869–877, 1992.
  • [STV01] Madhu Sudan, Luca Trevisan, and Salil P. Vadhan. Pseudorandom generators without the XOR lemma. J. Comput. Syst. Sci., 62(2):236–266, 2001.
  • [Zip79] Richard Zippel. Probabilistic algorithms for sparse polynomials. In Symbolic and Algebraic Computation, EUROSAM ’79, An International Symposiumon Symbolic and Algebraic Computation, Marseille, France, June 1979, Proceedings, pages 216–226, 1979.