arXiv subjects

B. Sundar Rajan

Publications and source records attributed to B. Sundar Rajan.

At least 91 records · Page 5Linked to original sources

arXiv2019

On the Optimality of Ali-Niesen Decentralized Coded Caching Scheme With and Without Error Correction

The decentralized coded caching was introduced in [M.~A. Maddah-Ali and U.~Niesen, `Decentralized coded caching attains order-optimal memory-rate tradeoff,' \emph{IEEE/ACM Trans. Networking}, vol.~23, no.~4, pp. 1029--1040, Aug. 2015] in which no coordination is required for the content placement. In this paper, with $N$ files and $K$ users, first it is shown that the decentralized coded caching scheme of Ali and Niesen is optimal for $N \geq K$ using techniques from index coding. Next the case when the shared link is error prone is considered. For this case, an optimal error correcting delivery scheme is proposed for coded caching problems with decentralized placement. Expression for the worst case rate is obtained when there are finite number of transmissions in error.

cs.IT↗

arXiv2019

On the Optimal Broadcast Rate of the Two-Sender Unicast Index Coding Problem with Fully-Participated Interactions

The problem of two-sender unicast index coding consists of two senders and a set of receivers. Each receiver demands a unique message and possesses some of the messages demanded by other receivers as its side-information. Every demanded message is present with at least one of the senders. Senders avail the knowledge of the side-information at the receivers to reduce the number of broadcast transmissions. Solution to this problem consists of finding the optimal number of coded transmissions from the two senders. One important class of the two-sender problem consists of the messages at the senders and the side-information at the receivers satisfying \emph{fully-participated interactions}. This paper provides the optimal broadcast rates, for all the unsolved cases of the two-sender problem with fully-participated interactions when the associated \emph{interaction digraphs} contain cycles. The optimal broadcast rates are provided in terms of those of the three independent single-sender problems associated with the two-sender problem. This paper also provides an achievable broadcast rate with $t$-bit messages for any finite $t$ and any two-sender problem with fully-participated interactions belonging to $(i)$ any one of the six instances (classes) of the two-sender problem when the associated interaction digraph does not contain any cycle, and $(ii)$ one of the classes of the two-sender problem when the associated interaction digraph contains cycles. The achievable broadcast rates are obtained by exploiting the symmetries of the confusion graph to color the same according to the two-sender graph coloring.

cs.IT↗

arXiv2019

A Generalisation of Interlinked Cycle Structures and Their Index Coding Capacity

Cycles and Cliques in a side-information graph reduce the number of transmissions required in an index coding problem. Thapa, Ong and Johnson defined a more general form of overlapping cycles, called the interlinked-cycle (IC) structure, that generalizes cycles and cliques. They proposed a scheme, that leverages IC structures in digraphs to construct scalar linear index codes. In this paper, we extend the notion of interlinked cycle structure to define more generalised graph structures called overlapping interlinked cycle (OIC) structures. We prove the capacity of OIC structures by giving an index code with length equal to the order of maximum acyclic induced subgraph (MAIS) of OIC structures.

cs.IT↗

arXiv2019

Reduced Complexity Index Codes and Improved Upperbound on Broadcast Rate for Neighboring Interference Problems

A single unicast index coding problem (SUICP) with symmetric neighboring interference (SNI) has $K$ messages and $K$ receivers, the $k$th receiver $R_{k}$ wanting the $k$th message $x_{k}$ and having the interference with $D$ messages after and $U$ messages before its desired message. Maleki, Cadambe and Jafar studied SUICP(SNI) because of its importance in topological interference management problems. Maleki \textit{et. al.} derived the lowerbound on the broadcast rate of this setting to be $D+1$. In our earlier work, for SUICP(SNI) with arbitrary $K,D$ and $U$, we defined set $\mathbf{S}$ of 2-tuples and for every $(a,b) \in \mathbf{S}$, we constructed $b$-dimensional vector linear index code with rate $D+1+\frac{a}{b}$ by using an encoding matrix of dimension $Kb \times (b(D+1)+a)$. In this paper, we use the symmetric structure of the SUICP(SNI) to reduce the size of encoding matrix by partitioning the message symbols. The rate achieved in this paper is same as that of the existing constructions of vector linear index codes. More specifically, we construct $b$-dimensional vector linear index codes for SUICP(SNI) by partitioning the $Kb$ messages into $b(U+1)+c$ sets for some non-negative integer $c$. We use an encoding matrix of size $\frac{Kb}{b(U+1)+c} \times \frac{b(D+1)+a}{b(U+1)+c}$ to encode each partition separately. The advantage of this method is that the receivers need to store atmost $\frac{b(D+1)+a}{b(U+1)+c}$ number of broadcast symbols (index code symbols) to decode a given wanted message symbol. We also give a construction of scalar linear index codes for SUICP(SNI) with arbitrary $K,D$ and $U$. We give an improved upperbound on the braodcast rate of SUICP(SNI).

cs.IT↗

arXiv2019

Groupcast Index Coding Problem: Joint Extensions

The groupcast index coding problem is the most general version of the classical index coding problem, where any receiver can demand messages that are also demanded by other receivers. Any groupcast index coding problem is described by its \emph{fitting matrix} which contains unknown entries along with $1$'s and $0$'s. The problem of finding an optimal scalar linear code is equivalent to completing this matrix with known entries such that the rank of the resulting matrix is minimized. Any row basis of such a completion gives an optimal \emph{scalar linear} code. An index coding problem is said to be a joint extension of a finite number of index coding problems, if the fitting matrices of these problems are disjoint submatrices of the fitting matrix of the jointly extended problem. In this paper, a class of joint extensions of any finite number of groupcast index coding problems is identified, where the relation between the fitting matrices of the sub-problems present in the fitting matrix of the jointly extended problem is defined by a base problem. A lower bound on the \emph{minrank} (optimal scalar linear codelength) of the jointly extended problem is given in terms of those of the sub-problems. This lower bound also has a dependence on the base problem and is operationally useful in finding lower bounds of the jointly extended problems when the minranks of all the sub-problems are known. We provide an algorithm to construct scalar linear codes (not optimal in general), for any groupcast problem belonging to the class of jointly extended problems identified in this paper. The algorithm uses scalar linear codes of all the sub-problems and the base problem. We also identify some subclasses, where the constructed codes are scalar linear optimal.

cs.IT↗

arXiv2019

A Relation Between Weight Enumerating Function and Number of Full Rank Sub-matrices

In most of the network coding problems with $k$ messages, the existence of binary network coding solution over $\mathbb{F}_2$ depends on the existence of adequate sets of $k$-dimensional binary vectors such that each set comprises of linearly independent vectors. In a given $k \times n$ ($n \geq k$) binary matrix, there exist ${n}\choose{k}$ binary sub-matrices of size $k \times k$. Every possible $k \times k$ sub-matrix may be of full rank or singular depending on the columns present in the matrix. In this work, for full rank binary matrix $\mathbf{G}$ of size $k \times n$ satisfying certain condition on minimum Hamming weight, we establish a relation between the number of full rank sub-matrices of size $k \times k$ and the weight enumerating function of the error correcting code with $\mathbf{G}$ as the generator matrix. We give an algorithm to compute the number of full rank $k \times k$ submatrices.

cs.IT↗

arXiv2019

On Pliable Index Coding

A new variant of index coding problem termed as Pliable Index Coding Problem (PICOD) is formulated in [S. Brahma, C. Fragouli, "Pliable index coding", IEEE Transactions on Information Theory, vol. 61, no. 11, pp. 6192-6203, 2015]. In PICOD, we consider a server holding a set of messages and there is a set of clients having a subset of messages with them. Each client is satisfied if it receives any of the message which it doesn't have. We discuss about a class of PICOD where the side information is consecutive. We provide index codes for two extreme cases - for the class where each client gets exactly one desired message and for a class where total number of messages decoded by the effective clients is maximized. Another variant of index coding problem is - c-Constrained Pliable Index Coding Problem [Linqi Song, Christina Fragouli and Tianchu Zhao, "A Pliable Index Coding Approach to Data Shuffling," arXiv:1701.05540v3 [cs.IT] 3 May 2018]. It is basically PICOD with a c-constraint, i.e, each message is decoded by atmost c clients demanding that message. We provide index codes for some classes of this variant with consecutive side information.

cs.IT↗

arXiv2019

An Optimal Linear Error Correcting Delivery Scheme for Coded Caching with Shared Caches

Classical coded caching setting avails each user to have one dedicated cache. This is generalized to a more general shared cache scheme and the exact expression for the worst case rate was derived in [E. Parrinello, A. Unsal, P. Elia, "Fundamental Limits of Caching in Heterogeneous Networks with Uncoded Prefetching," available on arXiv:1811.06247 [cs.IT], Nov. 2018]. For this case, an optimal linear error correcting delivery scheme is proposed and an expression for the peak rate is established for the same. Furthermore, a new delivery scheme is proposed, which gives an improved rate for the case when the demands are not distinct.

cs.IT↗

arXiv2019

A Field-Size Independent Code Construction for Groupcast Index Coding Problems

The length of an optimal scalar linear index code of a groupcast index coding problem is equal to the minrank of its side information hypergraph. The side-information hypergraph becomes a side-information graph for a special class of groupcast index coding problems known as unicast index coding problems. The computation of minrank is an NP-hard problem. There exists a low rank matrix completion method and clique cover method to find suboptimal solutions to the index coding problem represented by a side-information graph. However, both the methods are NP-hard. The number of computations required to find the minrank depends on the number of edges present in the side-information graph. In this paper, we define the notion of minrank-critical edges in a side-information graph and derive some properties of minrank, which identifies minrank-non-critical edges. Using these properties we present a method for reduction of the given minrank computation problem into a smaller problem. Also, we give an heuristic algorithm to find a clique cover of the side-information graph by using some binary operations on the adjacency matrix of the side-information graph. We also give a method to convert a groupcast index coding problem into a single unicast index coding problem. Combining all these results, we construct index codes (not necessarily optimal length) for groupcast index coding problems. The construction technique is independent of field size and hence can be used to construct index codes over binary field. In some cases the constructed index codes are better than the best known in the literature both in terms of the length of the code and the minimum field size required.

cs.IT↗

arXiv2018

Optimal Scalar Linear Index Codes for Symmetric and Neighboring Side-information Problems

A single unicast index coding problem (SUICP) is called symmetric neighboring and consecutive (SNC) side-information problem if it has $K$ messages and $K$ receivers, the $k$th receiver $R_{k}$ wanting the $k$th message $x_{k}$ and having the side-information $D$ messages immediately after $x_k$ and $U$ ($D\geq U$) messages immediately before $x_k$. Maleki, Cadambe and Jafar obtained the capacity of this SUICP(SNC) and proposed $(U+1)$-dimensional optimal length vector linear index codes by using Vandermonde matrices. However, for a $b$-dimensional vector linear index code, the transmitter needs to wait for $b$ realizations of each message and hence the latency introduced at the transmitter is proportional to $b$. For any given single unicast index coding problem (SUICP) with the side-information graph $G$, MAIS($G$) is used to give a lowerbound on the broadcast rate of the ICP. In this paper, we derive MAIS($G$) of SUICP(SNC) with side-information graph $G$. We construct scalar linear index codes for SUICP(SNC) with length $\left \lceil \frac{K}{U+1} \right \rceil - \left \lfloor \frac{D-U}{U+1} \right \rfloor$. We derive the minrank($G$) of SUICP(SNC) with side-information graph $G$ and show that the constructed scalar linear index codes are of optimal length for SUICP(SNC) with some combinations of $K,D$ and $U$. For SUICP(SNC) with arbitrary $K,D$ and $U$, we show that the length of constructed scalar linear index codes are atmost two index code symbols per message symbol more than the broadcast rate. The given results for SUICP(SNC) are of practical importance due to its relation with topological interference management problem in wireless communication networks.

cs.IT↗

arXiv2018

Optimal Linear Broadcast Rates of the Two-Sender Unicast Index Coding Problem with Fully-Participated Interactions

The two-sender unicast index coding problem consists of finding optimal coded transmissions from the two senders which collectively know the messages demanded by all the receivers. Each receiver demands a unique message. One important class of this problem consists of the message sets at the senders and the side-information at the receivers satisfying \emph{fully-participated interactions}. This paper provides optimal linear broadcast rates and corresponding code constructions for all the possible cases of the two-sender unicast index coding problem with fully-participated interactions. The optimal linear broadcast rate and the corresponding code for the two-sender problem are given in terms of those of the three single-sender unicast problems associated with the two-sender problem. Optimal linear broadcast rates of two-sender problems with fully-participated interactions provide lower bounds for the optimal linear broadcast rates of many related two-sender problems with \emph{partially-participated interactions}. Proof techniques used to obtain the results for the two-sender problem are shown to be useful in obtaining the results for some cases of the multi-sender unicast index coding problem.

cs.IT↗

arXiv2018

On the Structure of Interlinked Cycle Structures with Interlocked Outer Cycles

For index coding problems with special structure on the side-information graphs called Interlinked Cycle (IC) structures index codes have been proposed in the literature (C. Thapa, L. Ong, and S. Johnson, "Interlinked Cycles for Index Coding: Generalizing Cycles and Cliques", in IEEE Trans. Inf. Theory, vol. 63, no. 6, Jun. 2017, with a correction in "Interlinked Cycles for Index Coding: Generalizing Cycles and Cliques", in arxiv (arxiv:1603.00092v2 [cs.IT] 25 Feb 2018)). Recently (S. Sasi and B.S. Rajan, "On Optimal Index Codes for Interlinked Cycle Structures with Outer Cycles," in arxiv (arXiv:1804.09120v1 [cs.IT]), 24 Apr 2018) for a generalization of IC structures called IC structures with interlocked outer cycles optimal length index codes have been reported and it is shown that the optimal length depends on the maximum number of disjoint outer cycles. In this paper we discuss certain structural properties of IC structures with interlocked outer cycles and provide a simple algorithm to find the maximum number of disjoint outer cycles.

cs.IT↗

arXiv2018

On Optimal Index Codes for Interlinked Cycle Structures with Outer Cycles

For index coding problems with special structure on the side-information graphs called Interlinked Cycle (IC) structures index codes have been proposed in the literature (C. Thapa, L. Ong, and S. Johnson, "Interlinked Cycles for Index Coding: Generalizing Cycles and Cliques", in \textit{IEEE Trans. Inf. Theory, vol. 63, no. 6, Jun. 2017} with a correction in "Interlinked Cycles for Index Coding: Generalizing Cycles and Cliques", in arxiv (arxiv:1603.00092v2 [cs.IT] 25 Feb 2018)). In this paper we consider a generalization of IC structures called {\it IC structures with interlocked outer cycles}. For IC structures with interlocked outer cycles we show that the optimal length (also known as the minrank of the index coding problem) depends on the maximum number of disjoint outer cycles. We give two sufficient conditions such that if any of these is satisfied then we provide explicit optimal index code construction. The conditions mentioned above are shown to be not necessary by an explicit example.

cs.IT↗

arXiv2018

Index Codes for Interlinked Cycle Structures with Outer Cycles

Index code construction for a class of side-information graphs called interlinked cycle (IC) structures without outer cycles is given by Thapa, Ong and Johnson (C. Thapa, L. Ong, and S. Johnson, "Interlinked Cycles for Index Coding: Generalizing Cycles and Cliques", in \textit{IEEE Trans. Inf. Theory, vol. 63, no. 6, Jun. 2017}, "Interlinked Cycles for Index Coding: Generalizing Cycles and Cliques", in arxiv (arxiv:1603.00092v2 [cs.IT] 25 Feb 2018)). In this paper, construction of index codes for interlinked cycle (IC) structures with outer cycles is given along with a decoding algorithm.

cs.IT↗

arXiv2018

On Index Codes for Interlinked Cycle Structured Side-Information Graphs

In connection with the index code construction and the decoding algorithm for interlinked cycle (IC) structures proposed by Thapa, Ong and Johnson in \cite{TOJ} ("Interlinked Cycles for Index Coding: Generalizing Cycles and Cliques", IEEE Trans. Inf. Theory, vol. 63, no. 6, Jun. 2017), it is shown in \cite{VaR} ("Optimal Index Codes For A New Class of Interlinked Cycle Structure", in \textit{IEEE Communication Letters,} available as early access article in \textit{IEEE Xplore}: DOI-10.1109/LCOMM.2018.2799202) that the decoding algorithm does not work for all IC structures. In this work, a set of necessary and sufficient conditions on the IC structures is presented for the decoding algorithm to work for the code construction given in \cite{TOJ}. These conditions are shown to be satisfied for the IC structures without any cycles consisting of only non-inner vertices.

cs.IT↗

arXiv2018

On the Broadcast Rate of Index Coding Problems with Symmetric and Consecutive Interference

A single unicast index coding problem (SUICP) with symmetric and consecutive interference (SCI) has $K$ messages and $K$ receivers, the $k$th receiver $R_k$ wanting the $k$th message $x_k$ and having interference $\mathcal{I}_k= \{x_{k-U-m},\dots,x_{k-m-2},x_{k-m-1}\}\cup\{x_{k+m+1}, x_{k+m+2},\dots,x_{k+m+D}\}$ and side-information $\mathcal{K}_k=(\mathcal{I}_k \cup x_k)^c$. In this paper, we derive a lowerbound on the broadcast rate of single unicast index coding problem with symmetric and consecutive interference (SUICP(SCI)). In the SUICP(SCI), if $m=0$, we refer this as single unicast index coding problem with symmetric and neighboring interference (SUICP(SNI)). In our previous work\cite{VaR5}, we gave the construction of near-optimal vector linear index codes for SUICP(SNI) with arbitrary $K,D,U$. In this paper, we convert the SUICP(SCI) into SUICP(SNI) and give the construction of near-optimal vector linear index codes for SUICP(SCI) with arbitrary $K,U,D$ and $m$. The constructed codes are independent of field size. The near-optimal vector linear index codes of SUICP(SNI) is a special case of near-optimal vector linear index codes constructed in this paper for SUICP(SCI) with $m=0$. In our previous work\cite{VaR6}, we derived an upperbound on broadcast rate of SUICP(SNI). In this paper, we give an upperbound on the broadcast rate of SUICP(SCI) by using our earlier result on the upperbound on the broadcast rate of SUICP(SNI). We derive the capacity of SUICP(SCI) for some special cases.

cs.IT↗

arXiv2018

Optimal Error Correcting Delivery Scheme for an Optimal Coded Caching Scheme with Small Buffers

Optimal delivery scheme for coded caching problems with small buffer sizes and the number of users no less than the amount of files in the server was proposed by Chen, Fan and Letaief ["Fundamental limits of caching: improved bounds for users with small buffers," (IET Communications), 2016]. This scheme is referred to as the CFL scheme. In this paper, the link between the server and the users is assumed to be error prone only during the delivery phase. Closed form expressions for average rate and peak rate of error correcting delivery scheme for CFL prefetching scheme is obtained. An optimal error correcting delivery scheme for caching problems employing CFL prefetching is proposed.

cs.IT↗

arXiv2018

Reduced Dimensional Optimal Vector Linear Index Codes for Index Coding Problems with Symmetric Neighboring and Consecutive Side-information

A single unicast index coding problem (SUICP) with symmetric neighboring and consecutive side-information (SNCS) has $K$ messages and $K$ receivers, the $k$th receiver $R_k$ wanting the $k$th message $x_k$ and having the side-information $\mathcal{K}_k=\{x_{k-U},\dots,x_{k-2},x_{k-1}\}\cup\{x_{k+1}, x_{k+2},\dots,x_{k+D}\}$. The single unicast index coding problem with symmetric neighboring and consecutive side-information, SUICP(SNCS), is motivated by topological interference management problems in wireless communication networks. Maleki, Cadambe and Jafar obtained the symmetric capacity of this SUICP(SNCS) and proposed optimal length codes by using Vandermonde matrices. In our earlier work, we gave optimal length $(U+1)$-dimensional vector linear index codes for SUICP(SNCS) satisfying some conditions on $K,D$ and $U$ \cite{VaR1}. In this paper, for SUICP(SNCS) with arbitrary $K,D$ and $U$, we construct optimal length $\frac{U+1}{\text{gcd}(K,D-U,U+1)}$-dimensional vector linear index codes. We prove that the constructed vector linear index code is of minimal dimension if $\text{gcd}(K-D+U,U+1)$ is equal to $\text{gcd}(K,D-U,U+1)$. The proposed construction gives optimal length scalar linear index codes for the SUICP(SNCS) if $(U+1)$ divides both $K$ and $D-U$. The proposed construction is independent of field size and works over every field. We give a low-complexity decoding for the SUICP(SNCS). By using the proposed decoding method, every receiver is able to decode its wanted message symbol by simply adding some index code symbols (broadcast symbols).

cs.IT↗