Sparse Modeling Framework Research Articles

Hyperspectral image classification based on sparse modeling of spectral blocks

Hyperspectral images provide abundant spatial and spectral information that is very valuable for material detection in diverse areas of practical science. The high-dimensions of data lead to many processing challenges that can be addressed via existent spatial and spectral redundancies. In this paper, a sparse modeling framework is proposed for hyperspectral image classification. Spectral blocks are introduced to be used along with spatial groups to jointly exploit spectral and spatial redundancies. To reduce the computational complexity of sparse modeling, spectral blocks are used to break the high-dimensional optimization problems into small-size sub-problems that are faster to solve. Furthermore, the proposed sparse structure enables to extract the most discriminative spectral blocks and further reduce the computational burden. Experiments on three benchmark datasets, i.e., Pavia University, Indian Pines and Salinas images verify that the proposed method leads to a robust sparse modeling of hyperspectral images and improves the classification accuracy compared to several state-of-the-art methods. Moreover, the experiments demonstrate that the proposed method requires less processing time.

Power System Oscillation Mode Prediction Based on the Lasso Method

This paper utilizes modern statistical and machine learning methodology to predict the oscillation mode of interest in complex power engineering systems. The damping ratio of the electromechanical oscillation mode is formulated as a function of the power of the generators and loads as well as bus voltage magnitudes in the entire power system. The celebrated Lasso algorithm is implemented to solve this high-dimension modeling problem. By the nature of the L 1 design, the Lasso algorithm can automatically render a sparse solution, and by eliminating redundant features, it provides desirable prediction power. The resultant model processes a simple structure, and it is easily interpretable. The precision of our sparse modeling framework is demonstrated in the context of an IEEE 50-Generator 145-Bus power network and an online learning framework for the power system oscillation mode prediction is also provided.

High resolution sparse estimation of exponentially decaying N-dimensional signals

In this work, we consider the problem of high-resolution estimation of the parameters detailing an N-dimensional (N-D) signal consisting of an unknown number of exponentially decaying sinusoidal components. Since such signals are not sparse in an oversampled Fourier matrix, earlier approaches typically exploit large dictionary matrices that include not only a finely spaced frequency grid, but also a grid over the considered damping factors. Even in the 2-D case, the resulting dictionary is typically very large, resulting in a computationally cumbersome optimization problem. Here, we introduce a sparse modeling framework for N-dimensional exponentially damped sinusoids using the Kronecker structure inherent in the model. Furthermore, we introduce a novel dictionary learning approach that iteratively refines the estimate of the candidate frequency and damping coefficients for each component, thus allowing for smaller dictionaries, and for frequency and damping parameters that are not restricted to a grid. The performance of the proposed method is illustrated using simulated data, clearly showing the improved performance as compared to previous techniques.

Estimating Periodicities in Symbolic Sequences Using Sparse Modeling

In this paper, we propose a method for estimating statistical periodicities in symbolic sequences. Different from other common approaches used for the estimation of periodicities of sequences of arbitrary, finite, symbol sets, that often map the symbolic sequence to a numerical representation, we here exploit a likelihood-based formulation in a sparse modeling framework to represent the periodic behavior of the sequence. The resulting criterion includes a restriction on the cardinality of the solution; two approximate solutions are suggested-one greedy and one using an iterative convex relaxation strategy to ease the cardinality restriction. The performance of the proposed methods are illustrated using both simulated and real DNA data, showing a notable performance gain as compared to other common estimators.

Bayesian Generalized Kernel Mixed Models

We propose a fully Bayesian methodology for generalized kernel mixed models (GKMMs), which are extensions of generalized linear mixed models in the feature space induced by a reproducing kernel. We place a mixture of a point-mass distribution and Silverman's g-prior on the regression vector of a generalized kernel model (GKM). This mixture prior allows a fraction of the components of the regression vector to be zero. Thus, it serves for sparse modeling and is useful for Bayesian computation. In particular, we exploit data augmentation methodology to develop a Markov chain Monte Carlo (MCMC) algorithm in which the reversible jump method is used for model selection and a Bayesian model averaging method is used for posterior prediction. When the feature basis expansion in the reproducing kernel Hilbert space is treated as a stochastic process, this approach can be related to the Karhunen-Loeve expansion of a Gaussian process (GP). Thus, our sparse modeling framework leads to a flexible approximation method for GPs.