Approximate Subspace-Sparse Recovery in the Presence of Corruptions via `1-Minimization
نویسنده
چکیده
High-dimensional data often lie in low-dimensional subspaces corresponding to different classes they belong to. Finding sparse representations of data points in a dictionary built from the collection of data helps to uncover the low-dimensional subspaces and, as a result, address important problems such as compression, clustering, classification, subset selection and more. However, an important challenge related to real-world datasets is that the collection of data is often corrupted by measurement or process noise. In this paper, we address the problem of recovering sparse representations for noisy data points in a dictionary whose columns correspond to noisy data points lying close to a union of subspaces. We consider a constrained `1-minimization program and study conditions under which the solution of the optimization recovers a representation of a noisy point as a linear combination of a few noisy points from the same subspace. Our framework is based on a novel generalization of the null-space property to the setting where data lie in multiple subspaces, the number of data points in each subspace exceeds the dimension of the subspace, and all data points are corrupted by noise. We do not impose any randomness assumption on the arrangement of subspaces or distribution of data points in each subspace. We show that, under appropriate conditions that depend on relationships among data points within and between subspaces, the solution of our proposed optimization satisfies a desired approximate subspacesparse recovery. More specifically, we show that a noisy data point, close to one of the subspaces, will be reconstructed using data points from the same subspace with a small error and that the coefficients corresponding to data points in other subspaces will be sufficiently small.
منابع مشابه
Robust Subspace Recovery via Bi-Sparsity Pursuit
Successful applications of sparse models in computer vision and machine learning [3][2][5] imply that in many real-world applications, high dimensional data is distributed in a union of low dimensional subspaces. Nevertheless, the underlying structure may be affected by sparse errors and/or outliers. In this paper, we propose a dual sparse model as a framework to analyze this problem and provid...
متن کاملImage Classification via Sparse Representation and Subspace Alignment
Image representation is a crucial problem in image processing where there exist many low-level representations of image, i.e., SIFT, HOG and so on. But there is a missing link across low-level and high-level semantic representations. In fact, traditional machine learning approaches, e.g., non-negative matrix factorization, sparse representation and principle component analysis are employed to d...
متن کاملBian, Xiao. Sparse and Low-rank Modeling on High Dimensional Data: a Geometric Perspective. (under the Direction of Dr. Hamid Krim.) Sparse and Low-rank Modeling on High Dimensional Data: a Geometric Perspective
BIAN, XIAO. Sparse and Low-Rank Modeling on High Dimensional Data: A Geometric Perspective. (Under the direction of Dr. Hamid Krim.) High dimensional data exhibits distinct properties compared to its low dimensional counterpart, which causes a common performance decrease and a formidable computational cost increase of traditional approaches. Novel methodologies are therefore needed to character...
متن کاملMinimum error bounded efficient ℓ1 tracker with occlusion detection
Recently, sparse representation has been applied to visual tracking to find the target with the minimum reconstruction error from the target template subspace. Though effective, these L1 trackers require high computational costs due to numerous calculations for `1 minimization. In addition, the inherent occlusion insensitivity of the `1 minimization has not been fully utilized. In this paper, w...
متن کاملSubspace weighted l2, 1 minimization for sparse signal recovery
In this article, we propose a weighted l2,1 minimization algorithm for jointly-sparse signal recovery problem. The proposed algorithm exploits the relationship between the noise subspace and the overcomplete basis matrix for designing weights, i.e., large weights are appointed to the entries, whose indices are more likely to be outside of the row support of the jointly sparse signals, so that t...
متن کامل