arXiv subjects

Shuchin Aeron

Publications and source records attributed to Shuchin Aeron.

84 records · Page 5Linked to original sources

arXiv2014

First Order Methods for Robust Non-negative Matrix Factorization for Large Scale Noisy Data

Nonnegative matrix factorization (NMF) has been shown to be identifiable under the separability assumption, under which all the columns(or rows) of the input data matrix belong to the convex cone generated by only a few of these columns(or rows) [1]. In real applications, however, such separability assumption is hard to satisfy. Following [4] and [5], in this paper, we look at the Linear Programming (LP) based reformulation to locate the extreme rays of the convex cone but in a noisy setting. Furthermore, in order to deal with the large scale data, we employ First-Order Methods (FOM) to mitigate the computational complexity of LP, which primarily results from a large number of constraints. We show the performance of the algorithm on real and synthetic data sets.

stat.ML↗

arXiv2014

Robust Large Scale Non-negative Matrix Factorization using Proximal Point Algorithm

A robust algorithm for non-negative matrix factorization (NMF) is presented in this paper with the purpose of dealing with large-scale data, where the separability assumption is satisfied. In particular, we modify the Linear Programming (LP) algorithm of [9] by introducing a reduced set of constraints for exact NMF. In contrast to the previous approaches, the proposed algorithm does not require the knowledge of factorization rank (extreme rays [3] or topics [7]). Furthermore, motivated by a similar problem arising in the context of metabolic network analysis [13], we consider an entirely different regime where the number of extreme rays or topics can be much larger than the dimension of the data vectors. The performance of the algorithm for different synthetic data sets are provided.

stat.ML↗

arXiv2013

Consensus in the presence of interference

This paper studies distributed strategies for average-consensus of arbitrary vectors in the presence of network interference. We assume that the underlying communication on any \emph{link} suffers from \emph{additive interference} caused due to the communication by other agents following their own consensus protocol. Additionally, no agent knows how many or which agents are interfering with its communication. Clearly, the standard consensus protocol does not remain applicable in such scenarios. In this paper, we cast an algebraic structure over the interference and show that the standard protocol can be modified such that the average is reachable in a subspace whose dimension is complimentary to the maximal dimension of the interference subspaces (over all of the communication links). To develop the results, we use \emph{information alignment} to align the intended transmission (over each link) to the null-space of the interference (on that link). We show that this alignment is indeed invertible, i.e. the intended transmission can be recovered over which, subsequently, consensus protocol is implemented. That \emph{local} protocols exist even when the collection of the interference subspaces span the entire vector space is somewhat surprising.

eess.SY↗

arXiv2013

Joint multi-mode dispersion extraction in Fourier and space time domains

In this paper we present a novel broadband approach for the extraction of dispersion curves of multiple time frequency overlapped dispersive modes such as in borehole acoustic data. The new approach works jointly in the Fourier and space time domains and, in contrast to existing space time approaches that mainly work for time frequency separated signals, efficiently handles multiple signals with significant time frequency overlap. The proposed method begins by exploiting the slowness (phase and group) and time location estimates based on frequency-wavenumber (f-k) domain sparsity penalized broadband dispersion extraction method as presented in \cite{AeronTSP2011}. In this context we first present a Cramer Rao Bound (CRB) analysis for slowness estimation in the (f-k) domain and show that for the f-k domain broadband processing, group slowness estimates have more variance than the phase slowness estimates and time location estimates. In order to improve the group slowness estimates we exploit the time compactness property of the modes to effectively represent the data as a linear superposition of time compact space time propagators parameterized by the phase and group slowness. A linear least squares estimation algorithm in the space time domain is then used to obtain improved group slowness estimates. The performance of the method is demonstrated on real borehole acoustic data sets.

physics.geo-ph↗

arXiv2013

Novel Factorization Strategies for Higher Order Tensors: Implications for Compression and Recovery of Multi-linear Data

In this paper we propose novel methods for compression and recovery of multilinear data under limited sampling. We exploit the recently proposed tensor- Singular Value Decomposition (t-SVD)[1], which is a group theoretic framework for tensor decomposition. In contrast to popular existing tensor decomposition techniques such as higher-order SVD (HOSVD), t-SVD has optimality properties similar to the truncated SVD for matrices. Based on t-SVD, we first construct novel tensor-rank like measures to characterize informational and structural complexity of multilinear data. Following that we outline a complexity penalized algorithm for tensor completion from missing entries. As an application, 3-D and 4-D (color) video data compression and recovery are considered. We show that videos with linear camera motion can be represented more efficiently using t-SVD compared to traditional approaches based on vectorizing or flattening of the tensors. Application of the proposed tensor completion algorithm for video recovery from missing entries is shown to yield a superior performance over existing methods. In conclusion we point out several research directions and implications to online prediction of multilinear data.

cs.IT↗

arXiv2013

Methods for Large Scale Hydraulic Fracture Monitoring

In this paper we propose computationally efficient and robust methods for estimating the moment tensor and location of micro-seismic event(s) for large search volumes. Our contribution is two-fold. First, we propose a novel joint-complexity measure, namely the sum of nuclear norms which while imposing sparsity on the number of fractures (locations) over a large spatial volume, also captures the rank-1 nature of the induced wavefield pattern. This wavefield pattern is modeled as the outer-product of the source signature with the amplitude pattern across the receivers from a seismic source. A rank-1 factorization of the estimated wavefield pattern at each location can therefore be used to estimate the seismic moment tensor using the knowledge of the array geometry. In contrast to existing work this approach allows us to drop any other assumption on the source signature. Second, we exploit the recently proposed first-order incremental projection algorithms for a fast and efficient implementation of the resulting optimization problem and develop a hybrid stochastic & deterministic algorithm which results in significant computational savings.

physics.geo-ph↗

arXiv2013

Exploiting Structural Complexity for Robust and Rapid Hyperspectral Imaging

This paper presents several strategies for spectral de-noising of hyperspectral images and hypercube reconstruction from a limited number of tomographic measurements. In particular we show that the non-noisy spectral data, when stacked across the spectral dimension, exhibits low-rank. On the other hand, under the same representation, the spectral noise exhibits a banded structure. Motivated by this we show that the de-noised spectral data and the unknown spectral noise and the respective bands can be simultaneously estimated through the use of a low-rank and simultaneous sparse minimization operation without prior knowledge of the noisy bands. This result is novel for for hyperspectral imaging applications. In addition, we show that imaging for the Computed Tomography Imaging Systems (CTIS) can be improved under limited angle tomography by using low-rank penalization. For both of these cases we exploit the recent results in the theory of low-rank matrix completion using nuclear norm minimization.

physics.geo-ph↗

arXiv2013

Robust Hydraulic Fracture Monitoring (HFM) of Multiple Time Overlapping Events Using a Generalized Discrete Radon Transform

In this work we propose a novel algorithm for multiple-event localization for Hydraulic Fracture Monitoring (HFM) through the exploitation of the sparsity of the observed seismic signal when represented in a basis consisting of space time propagators. We provide explicit construction of these propagators using a forward model for wave propagation which depends non-linearly on the problem parameters - the unknown source location and mechanism of fracture, time and extent of event, and the locations of the receivers. Under fairly general assumptions and an appropriate discretization of these parameters we first build an over-complete dictionary of generalized Radon propagators and assume that the data is well represented as a linear superposition of these propagators. Exploiting this structure we propose sparsity penalized algorithms and workflow for super-resolution extraction of time overlapping multiple seismic events from single well data.

physics.geo-ph↗

arXiv2013

Complexity penalized hydraulic fracture localization and moment tensor estimation under limited model information

In this paper we present a novel technique for micro-seismic localization using a group sparse penalization that is robust to the focal mechanism of the source and requires only a velocity model of the stratigraphy rather than a full Green's function model of the earth's response. In this technique we construct a set of perfect delta detector responses, one for each detector in the array, to a seismic event at a given location and impose a group sparsity across the array. This scheme is independent of the moment tensor and exploits the time compactness of the incident seismic signal. Furthermore we present a method for improving the inversion of the moment tensor and Green's function when the geometry of seismic array is limited. In particular we demonstrate that both Tikhonov regularization and truncated SVD can improve the recovery of the moment tensor and be robust to noise. We evaluate our algorithm on synthetic data and present error bounds for both estimation of the moment tensor as well as localization. Furthermore we discuss the estimated moment tensor accuracy as a function of both array geometry and fault orientation.

physics.geo-ph↗

arXiv2010

Information theoretic bounds for Compressed Sensing

In this paper we derive information theoretic performance bounds to sensing and reconstruction of sparse phenomena from noisy projections. We consider two settings: output noise models where the noise enters after the projection and input noise models where the noise enters before the projection. We consider two types of distortion for reconstruction: support errors and mean-squared errors. Our goal is to relate the number of measurements, $m$, and $\snr$, to signal sparsity, $k$, distortion level, $d$, and signal dimension, $n$. We consider support errors in a worst-case setting. We employ different variations of Fano's inequality to derive necessary conditions on the number of measurements and $\snr$ required for exact reconstruction. To derive sufficient conditions we develop new insights on max-likelihood analysis based on a novel superposition property. In particular this property implies that small support errors are the dominant error events. Consequently, our ML analysis does not suffer the conservatism of the union bound and leads to a tighter analysis of max-likelihood. These results provide order-wise tight bounds. For output noise models we show that asymptotically an $\snr$ of $Θ(\log(n))$ together with $Θ(k \log(n/k))$ measurements is necessary and sufficient for exact support recovery. Furthermore, if a small fraction of support errors can be tolerated, a constant $\snr$ turns out to be sufficient in the linear sparsity regime. In contrast for input noise models we show that support recovery fails if the number of measurements scales as $o(n\log(n)/SNR)$ implying poor compression performance for such cases. We also consider Bayesian set-up and characterize tradeoffs between mean-squared distortion and the number of measurements using rate-distortion theory.

cs.IT↗

arXiv2007

On sensing capacity of sensor networks for the class of linear observation, fixed SNR models

In this paper we address the problem of finding the sensing capacity of sensor networks for a class of linear observation models and a fixed SNR regime. Sensing capacity is defined as the maximum number of signal dimensions reliably identified per sensor observation. In this context sparsity of the phenomena is a key feature that determines sensing capacity. Precluding the SNR of the environment the effect of sparsity on the number of measurements required for accurate reconstruction of a sparse phenomena has been widely dealt with under compressed sensing. Nevertheless the development there was motivated from an algorithmic perspective. In this paper our aim is to derive these bounds in an information theoretic set-up and thus provide algorithm independent conditions for reliable reconstruction of sparse signals. In this direction we first generalize the Fano's inequality and provide lower bounds to the probability of error in reconstruction subject to an arbitrary distortion criteria. Using these lower bounds to the probability of error, we derive upper bounds to sensing capacity and show that for fixed SNR regime sensing capacity goes down to zero as sparsity goes down to zero. This means that disproportionately more sensors are required to monitor very sparse events. Our next main contribution is that we show the effect of sensing diversity on sensing capacity, an effect that has not been considered before. Sensing diversity is related to the effective \emph{coverage} of a sensor with respect to the field. In this direction we show the following results (a) Sensing capacity goes down as sensing diversity per sensor goes down; (b) Random sampling (coverage) of the field by sensors is better than contiguous location sampling (coverage).

cs.IT↗

arXiv2006

Wireless ad-hoc networks: Strategies and Scaling laws for the fixed SNR regime

This paper deals with throughput scaling laws for random ad-hoc wireless networks in a rich scattering environment. We develop schemes to optimize the ratio, $ρ(n)$ of achievable network sum capacity to the sum of the point-to-point capacities of source-destinations pairs operating in isolation. For fixed SNR networks, i.e., where the worst case SNR over the source-destination pairs is fixed independent of $n$, we show that collaborative strategies yield a scaling law of $ρ(n) = {\cal O}(\frac{1}{n^{1/3}})$ in contrast to multi-hop strategies which yield a scaling law of $ρ(n) = {\cal O}(\frac{1}{\sqrt{n}})$. While, networks where worst case SNR goes to zero, do not preclude the possibility of collaboration, multi-hop strategies achieve optimal throughput. The plausible reason is that the gains due to collaboration cannot offset the effect of vanishing receive SNR. This suggests that for fixed SNR networks, a network designer should look for network protocols that exploit collaboration. The fact that most current networks operate in a fixed SNR interference limited environment provides further motivation for considering this regime.

cs.IT↗