arXiv subjects

C. S. Sastry

Publications and source records attributed to C. S. Sastry.

9 recordsLinked to original sources

arXiv2021

Nullspace Property for Optimality of Minimum Frame Angle Under Invertible Linear Operators

Orthogonal Matching Pursuit and Basis Pursuit are popular reconstruction algorithms for recovery of sparse signals. The exact recovery property of both the methods has a relation with the coherence of the underlying redundant dictionary, i.e. a frame. A frame with low coherence provides better guarantees for exact recovery. An equivalent formulation of the associated linear system is obtained via premultiplication by a non-singular matrix. In view of bounds that guarantee sparse recovery, it is very useful to generate the preconditioner in such way that the preconditioned frame has low coherence as compared to the original. In this paper, we discuss the impact of preconditioning on sparse recovery. Further, we formulate a convex optimization problem for designing the preconditioner that yields a frame with improved coherence. In addition to reducing coherence, we focus on designing well conditioned frames and numerically study the relationship between the condition number of the preconditioner and the coherence of the new frame. Alongside theoretical justifications, we demonstrate through simulations the efficacy of the preconditioner in reducing coherence as well as recovering sparse signals.

math.OC↗

arXiv2020

On compactly supported discrete radial wavelets in $L^2(\mathbb{R}^2)$ and application in Tomography

Radially symmetric wavelets possessing multiresolution framework are found to be useful in different fields like Pattern recognition, Computed Tomography (CT) etc. The compactly supported wavelets are known to be useful for localized operations in applications such as reconstruction, enhancement etc. In this work we introduce a novel way of designing compactly supported radial wavelets in $L^2(\mathbb{R}^2)$ from a 1D Daubechies wavelets and obtain a reconstruction formula possessing multiresolution framework. Further, we demonstrate the usefulness of our radial wavelets in Tomography.

math.FA↗

arXiv2020

Local recovery bounds for prior support constrained Compressed Sensing

Prior support constrained compressed sensing has of late become popular due to its potential for applications. The existing results on recovery guarantees provide global recovery bounds in the sense that they deal with full support. However, in some applications, one might be interested in the recovery guarantees limited to the given prior support, such bounds may be termed as local recovery bounds. The present work proposes the local recovery guarantees and analyzes the conditions on associated parameters that make recovery error small.

math.OC↗

arXiv2019

Sufficient conditions for the uniqueness of solution of the weighted norm minimization problem

Prior support constrained compressed sensing, achieved via the weighted norm minimization, has of late become popular due to its potential for applications. For the weighted norm minimization problem, $$ min \|x\|_{p,w} \text{ subject to } y=Ax, \; p=0,1, \text{ and } w \in [0,1], $$ uniqueness results are known when $w=0,1$. Here, $\|x\|_{p,w}=w\|x_T\|_p+\|x_{T^c}\|_p, \; p=0,1$ with $T$ representing the partial support information. The work reported in this paper presents the conditions that ensure the uniqueness of the solution of this problem for general $w \in [0,1]$.

math.OC↗

arXiv2019

Sparse recovery guarantees for block orthogonal binary matrices constructed via Generalized Euler Squares

In recent times, the construction of deterministic matrices has gained popularity as an alternative of random matrices as they provide guarantees for recovery of sparse signals. In particular, the construction of binary matrices has attained significance due to their potential for hardware-friendly implementation and appealing applications. Our present work aims at constructing incoherent binary matrices consisting of orthogonal blocks with small block coherence. We show that the binary matrices constructed from Euler squares exhibit block orthogonality and possess low block coherence. With a goal of obtaining better aspect ratios, the present work generalizes the notion of Euler Squares and obtains a new class of deterministic binary matrices of more general size. For realizing the stated objectives, to begin with, the paper revisits the connection of finite field theory to Euler Squares and their construction. Using the stated connection, the work proposes Generalized Euler Squares (GES) and then presents a construction procedure. Binary matrices with low coherence and general row-sizes are obtained, whose column size is in the maximum possible order. Finally, the paper shows that the special structure possessed by GES is helpful in resulting in block orthogonal structure with small block coherence, which supports the recovery of block sparse signals.

math.CO↗

arXiv2018

Novel Light Weight Compressed Data Aggregation Using Sparse Measurements for IoT Networks

Optimal data aggregation aimed at maximizing IoT network lifetime by minimizing constrained on-board resource utilization continues to be a challenging task. The existing data aggregation methods have proven that compressed sensing is promising for data aggregation. However, they compromise either on energy efficiency or recovery fidelity and require complex on-node computations. In this paper, we propose a novel Light Weight Compressed Data Aggregation (LWCDA) algorithm that randomly divides the entire network into non-overlapping clusters for data aggregation. The random non-overlapping clustering offers two important advantages: 1) energy efficiency, as each node has to send its measurement only to its cluster head, 2) highly sparse measurement matrix, which leads to a practically implementable framework with low complexity. We analyze the properties of our measurement matrix using restricted isometry property, the associated coherence and phase transition. Through extensive simulations on practical data, we show that the measurement matrix can reconstruct data with high fidelity. Further, we demonstrate that the LWCDA algorithm reduces transmission cost significantly against baseline approaches, implying thereby the enhancement of the network lifetime.

eess.SP↗

arXiv2017

Construction of Structured Incoherent Unit Norm Tight Frames

The exact recovery property of Basis pursuit (BP) and Orthogonal Matching Pursuit (OMP) has a relation with the coherence of the underlying frame. A frame with low coherence provides better guarantees for exact recovery. In particular, Incoherent Unit Norm Tight Frames (IUNTFs) play a significant role in sparse representations. IUNTFs with special structure, in particular those given by a union of several orthonormal bases, are known to satisfy better theoretical guarantees for recovering sparse signals. In the present work, we propose to construct structured IUNTFs consisting of large number of orthonormal bases. For a given $r, k, m$ with $k$ being less than or equal to the smallest prime power factor of $m$ and $r<k,$ we construct a CS matrix of size $mk \times (mk\times m^{r})$ with coherence at most $\frac{r}{k},$ which consists of $m^{r}$ number of orthonormal bases and with density $\frac{1}{m}$. We also present numerical results of recovery performance of union of orthonormal bases as against their Gaussian counterparts.

math.NA↗

arXiv2016

Deterministic compressed sensing matrices: Construction via Euler Squares and applications

In Compressed Sensing the matrices that satisfy the Restricted Isometry Property (RIP) play an important role. But to date, very few results for designing such matrices are available. For applications such as multiplier-less data compression, binary sensing matrices are of interest. The present work constructs deterministic and binary sensing matrices using Euler Squares. In particular, given a positive integer $m$ different from $p, p^2$ for a prime $p$, we show that it is possible to construct a binary sensing matrix of size $m \times c (mμ)^2$, where $μ$ is the coherence parameter of the matrix and $c \in [1,2)$. The matrices that we construct have smaller density (that is, percentage of nonzero entries in the matrix is small) with no function evaluation in their construction, which support algorithms with low computational complexity. Through experimental work, we show that our binary sensing matrices can be used for such applications as content based image retrieval. Our simulation results demonstrate that the Euler Square based CS matrices give better performance than their Gaussian counterparts.

math.OC↗

arXiv2015

Deterministic construction of sparse binary and ternary matrices from existing binary sensing matrices

In the present work, we discuss a procedure for constructing sparse binary and ternary matrices from existing two binary sensing matrices. The matrices that we construct have several attractive properties such as smaller density, which supports algorithms with low computational complexity. As an application of our method, we show that a CS matrix of general row size different from $p, p^2, pq$ (for different primes $p,q$) can be constructed.

cs.IT↗