Results 11  20
of
325
Quantitative Robust Uncertainty Principles and Optimally Sparse Decompositions
, 2004
"... In this paper, we develop a robust uncertainty principle for finite signals in C N which states that for nearly all choices T, Ω ⊂ {0,..., N − 1} such that T  + Ω  ≍ (log N) −1/2 · N, there is no signal f supported on T whose discrete Fourier transform ˆ f is supported on Ω. In fact, we can mak ..."
Abstract

Cited by 119 (12 self)
 Add to MetaCart
In this paper, we develop a robust uncertainty principle for finite signals in C N which states that for nearly all choices T, Ω ⊂ {0,..., N − 1} such that T  + Ω  ≍ (log N) −1/2 · N, there is no signal f supported on T whose discrete Fourier transform ˆ f is supported on Ω. In fact, we can make the above uncertainty principle quantitative in the sense that if f is supported on T, then only a small percentage of the energy (less than half, say) of ˆ f is concentrated on Ω. As an application of this robust uncertainty principle (QRUP), we consider the problem of decomposing a signal into a sparse superposition of spikes and complex sinusoids f(s) = � α1(t)δ(s − t) + � α2(ω)e i2πωs/N / √ N. t∈T We show that if a generic signal f has a decomposition (α1, α2) using spike and frequency locations in T and Ω respectively, and obeying ω∈Ω T  + Ω  ≤ Const · (log N) −1/2 · N, then (α1, α2) is the unique sparsest possible decomposition (all other decompositions have more nonzero terms). In addition, if T  + Ω  ≤ Const · (log N) −1 · N, then the sparsest (α1, α2) can be found by solving a convex optimization problem. Underlying our results is a new probabilistic approach which insists on finding the correct uncertainty relation or the optimally sparse solution for nearly all subsets but not necessarily all of them, and allows to considerably sharpen previously known results [9, 10]. In fact, we show that the fraction of sets (T, Ω) for which the above properties do not hold can be upper bounded by quantities like N −α for large values of α. The QRUP (and the application to finding sparse representations) can be extended to general pairs of orthogonal bases Φ1, Φ2 of C N. For nearly all choices Γ1, Γ2 ⊂ {0,..., N − 1} obeying Γ1  + Γ2  ≍ µ(Φ1, Φ2) −2 · (log N) −m, where m ≤ 6, there is no signal f such that Φ1f is supported on Γ1 and Φ2f is supported on Γ2 where µ(Φ1, Φ2) is the mutual coherence between Φ1 and Φ2.
On sparse reconstruction from Fourier and Gaussian measurements
 Communications on Pure and Applied Mathematics
, 2006
"... Abstract. This paper improves upon best known guarantees for exact reconstruction of a sparse signal f from a small universal sample of Fourier measurements. The method for reconstruction that has recently gained momentum in the Sparse Approximation Theory is to relax this highly nonconvex problem ..."
Abstract

Cited by 116 (9 self)
 Add to MetaCart
Abstract. This paper improves upon best known guarantees for exact reconstruction of a sparse signal f from a small universal sample of Fourier measurements. The method for reconstruction that has recently gained momentum in the Sparse Approximation Theory is to relax this highly nonconvex problem to a convex problem, and then solve it as a linear program. We show that there exists a set of frequencies Ω such that one can exactly reconstruct every rsparse signal f of length n from its frequencies in Ω, using the convex relaxation, and Ω has size k(r, n) = O(r log(n)·log 2 (r) log(r log n)) = O(r log 4 n). A random set Ω satisfies this with high probability. This estimate is optimal within the log log n and log 3 r factors. We also give a relatively short argument for a similar problem with k(r, n) � r[12 + 8 log(n/r)] Gaussian measurements. We use methods of geometric functional analysis and probability theory in Banach spaces, which makes our arguments quite short. 1.
Sparse multinomial logistic regression: fast algorithms and generalization bounds
 IEEE Trans. on Pattern Analysis and Machine Intelligence
"... Abstract—Recently developed methods for learning sparse classifiers are among the stateoftheart in supervised learning. These methods learn classifiers that incorporate weighted sums of basis functions with sparsitypromoting priors encouraging the weight estimates to be either significantly larg ..."
Abstract

Cited by 113 (1 self)
 Add to MetaCart
Abstract—Recently developed methods for learning sparse classifiers are among the stateoftheart in supervised learning. These methods learn classifiers that incorporate weighted sums of basis functions with sparsitypromoting priors encouraging the weight estimates to be either significantly large or exactly zero. From a learningtheoretic perspective, these methods control the capacity of the learned classifier by minimizing the number of basis functions used, resulting in better generalization. This paper presents three contributions related to learning sparse classifiers. First, we introduce a true multiclass formulation based on multinomial logistic regression. Second, by combining a bound optimization approach with a componentwise update procedure, we derive fast exact algorithms for learning sparse multiclass classifiers that scale favorably in both the number of training samples and the feature dimensionality, making them applicable even to large data sets in highdimensional feature spaces. To the best of our knowledge, these are the first algorithms to perform exact multinomial logistic regression with a sparsitypromoting prior. Third, we show how nontrivial generalization bounds can be derived for our classifier in the binary case. Experimental results on standard benchmark data sets attest to the accuracy, sparsity, and efficiency of the proposed methods.
Geometric approach to error correcting codes and reconstruction of signals
 INT. MATH. RES. NOT
, 2005
"... ..."
Neighborly Polytopes and Sparse Solutions of Underdetermined Linear Equations
, 2005
"... Consider a d × n matrix A, with d < n. The problem of solving for x in y = Ax is underdetermined, and has many possible solutions (if there are any). In several fields it is of interest to find the sparsest solution – the one with fewest nonzeros – but in general this involves combinatorial optimiza ..."
Abstract

Cited by 85 (12 self)
 Add to MetaCart
Consider a d × n matrix A, with d < n. The problem of solving for x in y = Ax is underdetermined, and has many possible solutions (if there are any). In several fields it is of interest to find the sparsest solution – the one with fewest nonzeros – but in general this involves combinatorial optimization. Let ai denote the ith column of A, 1 ≤ i ≤ n. Associate to A the quotient polytope P formed by taking the convex hull of the 2n points (±ai) in R d. P is centrosymmetric and is called (centrally) kneighborly if every subset of k + 1 elements (±ilail)k+1 l=1 are the vertices of a face of P. We show that if P is kneighborly, then if a system y = Ax has a solution with at most k nonzeros, that solution is also the unique solution of the convex optimization problem min �x�1 subject to y = Ax; the converse holds as well. This complete equivalence between the study of sparse solution by ℓ 1 minimization and neighborliness of convex polytopes immediately gives new results in each field. On the one
Sparse reconstruction by convex relaxation: Fourier and Gaussian measurements
 CISS 2006 (40th Annual Conference on Information Sciences and Systems
, 2006
"... Abstract — This paper proves best known guarantees for exact reconstruction of a sparse signal f from few nonadaptive universal linear measurements. We consider Fourier measurements (random sample of frequencies of f) and random Gaussian measurements. The method for reconstruction that has recently ..."
Abstract

Cited by 81 (5 self)
 Add to MetaCart
Abstract — This paper proves best known guarantees for exact reconstruction of a sparse signal f from few nonadaptive universal linear measurements. We consider Fourier measurements (random sample of frequencies of f) and random Gaussian measurements. The method for reconstruction that has recently gained momentum in the Sparse Approximation Theory is to relax this highly nonconvex problem to a convex problem, and then solve it as a linear program. What are best guarantees for the reconstruction problem to be equivalent to its convex relaxation is an open question. Recent work shows that the number of measurements k(r, n) needed to exactly reconstruct any rsparse signal f of length n from its linear measurements with convex relaxation is usually O(r polylog(n)). However, known guarantees involve huge constants, in spite of very good performance of the algorithms in practice. In attempt to reconcile theory with practice, we prove the first guarantees for universal measurements (i.e. which work for all sparse functions) with reasonable constants. For Gaussian measurements, k(r, n) � 11.7 r ˆ 1.5 + log(n/r) ˜ , which is optimal up to constants. For Fourier measurements, we prove the best known bound k(r, n) = O(r log(n) · log 2 (r) log(r log n)), which is optimal within the log log n and log 3 r factors. Our arguments are based on the
Ranksparsity incoherence for matrix decomposition
, 2009
"... Abstract. Suppose we are given a matrix that is formed by adding an unknown sparse matrix to an unknown lowrank matrix. Our goal is to decompose the given matrix into its sparse and lowrank components. Such a problem arises in a number of applications in model and system identification, and is int ..."
Abstract

Cited by 81 (10 self)
 Add to MetaCart
Abstract. Suppose we are given a matrix that is formed by adding an unknown sparse matrix to an unknown lowrank matrix. Our goal is to decompose the given matrix into its sparse and lowrank components. Such a problem arises in a number of applications in model and system identification, and is intractable to solve in general. In this paper we consider a convex optimization formulation to splitting the specified matrix into its components, by minimizing a linear combination of the ℓ1 norm and the nuclear norm of the components. We develop a notion of ranksparsity incoherence, expressed as an uncertainty principle between the sparsity pattern of a matrix and its row and column spaces, and use it to characterize both fundamental identifiability as well as (deterministic) sufficient conditions for exact recovery. Our analysis is geometric in nature with the tangent spaces to the algebraic varieties of sparse and lowrank matrices playing a prominent role. When the sparse and lowrank matrices are drawn from certain natural random ensembles, we show that the sufficient conditions for exact recovery are satisfied with high probability. We conclude with simulation results on synthetic matrix decomposition problems.
Sparsest solutions of underdetermined linear systems via ℓ
"... We present a condition on the matrix of an underdetermined linear system which guarantees that the solution of the system with minimal ℓqquasinorm is also the sparsest one. This generalizes, and sightly improves, a similar result for the ℓ1norm. We then introduce a simple numerical scheme to compu ..."
Abstract

Cited by 77 (8 self)
 Add to MetaCart
We present a condition on the matrix of an underdetermined linear system which guarantees that the solution of the system with minimal ℓqquasinorm is also the sparsest one. This generalizes, and sightly improves, a similar result for the ℓ1norm. We then introduce a simple numerical scheme to compute solutions with minimal ℓqquasinorm, and we study its convergence. Finally, we display the results of some experiments which indicate that the ℓqmethod performs better than other available methods. 1
Sparse Bayesian learning for basis selection
 IEEE Transactions on Signal Processing
, 2004
"... Abstract—Sparse Bayesian learning (SBL) and specifically relevance vector machines have received much attention in the machine learning literature as a means of achieving parsimonious representations in the context of regression and classification. The methodology relies on a parameterized prior tha ..."
Abstract

Cited by 75 (5 self)
 Add to MetaCart
Abstract—Sparse Bayesian learning (SBL) and specifically relevance vector machines have received much attention in the machine learning literature as a means of achieving parsimonious representations in the context of regression and classification. The methodology relies on a parameterized prior that encourages models with few nonzero weights. In this paper, we adapt SBL to the signal processing problem of basis selection from overcomplete dictionaries, proving several results about the SBL cost function that elucidate its general behavior and provide solid theoretical justification for this application. Specifically, we have shown that SBL retains a desirable property of the 0norm diversity measure (i.e., the global minimum is achieved at the maximally sparse solution) while often possessing a more limited constellation of local minima. We have also demonstrated that the local minima that do exist are achieved at sparse solutions. Later, we provide a novel interpretation of SBL that gives us valuable insight into why it is successful in producing sparse representations. Finally, we include simulation studies comparing sparse Bayesian learning with Basis Pursuit and the more recent FOCal Underdetermined System Solver (FOCUSS) class of basis selection algorithms. These results indicate that our theoretical insights translate directly into improved performance. Index Terms—Basis selection, diversity measures, linear inverse problems, sparse Bayesian learning, sparse representations. I.
Analysis versus synthesis in signal priors
, 2005
"... The concept of prior probability for signals plays a key role in the successful solution of many inverse problems. Much of the literature on this topic can be divided between analysisbased and synthesisbased priors. Analysisbased priors assign probability to a signal through various forward measu ..."
Abstract

Cited by 69 (12 self)
 Add to MetaCart
The concept of prior probability for signals plays a key role in the successful solution of many inverse problems. Much of the literature on this topic can be divided between analysisbased and synthesisbased priors. Analysisbased priors assign probability to a signal through various forward measurements of it, while synthesisbased priors seek a reconstruction of the signal as a combination of atom signals. In this paper we describe these two prior classes, focusing on the distinction between them. We show that although when reducing to the complete and undercomplete formulations the two become equivalent, in their more interesting overcomplete formulation the two types depart. Focusing on the ℓ1 denoising case, we present several ways of comparing the two types of priors, establishing the existence of an unbridgeable gap between them. 1.