This paper describes our submission to the eighth annual mlsp competition organized by Amazon during the 2012 ieeemlspworkshop. Our approach is based on a nearest-neighbor-like classifier with a distance metric lear...
ISBN:
(纸本)9781467310260
This paper describes our submission to the eighth annual mlsp competition organized by Amazon during the 2012 ieeemlspworkshop. Our approach is based on a nearest-neighbor-like classifier with a distance metric learned from samples. The method was second in the final standings with prediction accuracy of 81 %, while the winning submission was 87 % accurate.
We propose a new framework for Detection/Estimation designed to avoid the loss of salient information in the process of reducing the dimensionality of digitized data. The main idea is a Semi-Supervised learning pre-pr...
详细信息
ISBN:
(纸本)9781424423750
We propose a new framework for Detection/Estimation designed to avoid the loss of salient information in the process of reducing the dimensionality of digitized data. The main idea is a Semi-Supervised learning pre-processing scheme ([1]) based oil Compressed Sensing ([2]). The proposed approach combines a first step -performed at the data acquisition level- with an energy based algorithm ([3], [4]) aimed at defining a global metric on the data. The latter is then used to drive the classification algorithm. We demonstrate the power of the new technique by applying it to the detection of cellular nuclei in large, high-dimensional, hyperspectral images.
This paper presents anew blind signal extraction method based on mutual information. Conventional blind signal extraction methods minimize the mutual information between the extracted signal and the remaining signals ...
详细信息
ISBN:
(纸本)9781424423750
This paper presents anew blind signal extraction method based on mutual information. Conventional blind signal extraction methods minimize the mutual information between the extracted signal and the remaining signals indirectly by using a cost function. The proposed method directly minimizes this mutual information through a gradient descent. The derivation of the gradient exploits recent results on the differential of the mutual information and the implementation is based on kernel based density estimation. Some simulation results show the performance of the proposed approach and underline the improvement obtained by using the proposed method as a post-processing for conventional methods.
A variational approach based on level set methods popular in image segmentation is presented for learning discriminative classifiers in general feature spaces. Nonlinear, nonparametric decision boundaries are obtained...
详细信息
ISBN:
(纸本)9781424423750
A variational approach based on level set methods popular in image segmentation is presented for learning discriminative classifiers in general feature spaces. Nonlinear, nonparametric decision boundaries are obtained by minimizing an energy functional that incorporates a margin-based loss function. The class of level set contour decision boundaries is discussed in terms of the structural risk minimization principle. A variation on El feature subset selection is developed. Use of level set classifiers as base learners for boosting is discussed.
Dimensionality reduction is a well known technique in signalprocessing oriented to improve both the computational cost and the performance of classifiers. We use an electroencephalogram (EEG) feature matrix based on ...
详细信息
ISBN:
(纸本)9781457716232
Dimensionality reduction is a well known technique in signalprocessing oriented to improve both the computational cost and the performance of classifiers. We use an electroencephalogram (EEG) feature matrix based on three extraction methods: tracks extraction, wavelets coefficients and Fractional Fourier Transform. The dimension reduction is performed by Mutual Information (MI) and a forward-backward procedure. Our results show that feature extraction and dimension reduction could be considered as a new alternative for solving EEG classification problems.
Atrial fibrillation (AF) is a common heart disorder. One of the most prominent hypothesis about its initiation and maintenance considers multiple uncoordinated activation foci inside the atrium. However, the implicit ...
详细信息
ISBN:
(纸本)9781467310260
Atrial fibrillation (AF) is a common heart disorder. One of the most prominent hypothesis about its initiation and maintenance considers multiple uncoordinated activation foci inside the atrium. However, the implicit assumption behind all the signalprocessing techniques used for AF, such as dominant frequency and organization analysis, is the existence of a single regular component in the observed signals. In this paper we take into account the existence of multiple foci, performing a spectral analysis to detect their number and frequencies. In order to obtain a cleaner signal on which the spectral analysis can be performed, we introduce sparsity-aware learning techniques to infer the spike trains corresponding to the activations. The good performance of the proposed algorithm is demonstrated both on synthetic and real data.
In this paper, we propose a two-tier method for extracting fetal ECG from a single lead abdominal ECG signal. The proposed method is based on a combination of singular value decomposition (SVD) and polynomial classifi...
详细信息
ISBN:
(纸本)9781424423750
In this paper, we propose a two-tier method for extracting fetal ECG from a single lead abdominal ECG signal. The proposed method is based on a combination of singular value decomposition (SVD) and polynomial classifiers. As a first tier, SVD is used to extract an estimate of the maternal component from the composite abdominal signal by exploiting its quasi-periodic nature. The extracted maternal signal is then used along with the abdominal composite signal to isolate the FECG component using polynomial classifiers. The proposed method is validated on both real and synthetic data. Results demonstrate effectiveness of proposed method.
One common problem in signal denoising is that if the signal has a blocky, in other words a piecewise-smooth structure, the denoised signal may suffer from oversmoothed discontinuities or exhibit artifacts very simila...
详细信息
ISBN:
(纸本)9781424423750
One common problem in signal denoising is that if the signal has a blocky, in other words a piecewise-smooth structure, the denoised signal may suffer from oversmoothed discontinuities or exhibit artifacts very similar to Gibbs phenomenon. In the literature, total variation methods and some modifications on the signal reconstructions based on wavelet coefficients are proposed to overcome these problems. We take a novel approach by introducing principal curve projections as an artifact-free signal denoising filter alternative. The proposed approach leads to a nonparametric denoising algorithm that does not lead to Gibbs effect or so-called staircase type unnatural artifacts in the denoised signal.
Maximum Variance Unfolding (MVU) is among the state of the art Manifold learning (ML) algorithms and experimentally proven to be the best method to unfold a manifold to its intrinsic dimension. Unfortunately it doesn&...
详细信息
ISBN:
(纸本)9781424423750
Maximum Variance Unfolding (MVU) is among the state of the art Manifold learning (ML) algorithms and experimentally proven to be the best method to unfold a manifold to its intrinsic dimension. Unfortunately it doesn't scale for more than a few hundred points. A non convex formulation of MVU made it possible to scale up to a few thousand points with the risk of getting trapped in local minima. In this paper we demonstrate techniques based on the dual-tree algorithm and L-BFGS that allow MVU to scale up to 100,000 points. We also present a new variant called Maximum Furthest Neighbor Unfolding (MFNU) which performs even better than MVU in terms of avoiding local minima.
Adaptive filters are crucial in many signalprocessing applications. Recently, a simple configuration was presented to introduce a bias in the estimation of adaptive filters using a multiplicative factor, showing impo...
详细信息
ISBN:
(纸本)9781457716232
Adaptive filters are crucial in many signalprocessing applications. Recently, a simple configuration was presented to introduce a bias in the estimation of adaptive filters using a multiplicative factor, showing important gains in terms of mean square error with respect to standard adaptive filter operation, mainly for low signal to noise ratios. In this paper, we modify that scheme to obtain further advantages by splitting the adaptive filter coefficients into non-overlapping blocks, and employing a different multiplicative factor for the coefficients in each block. In this way, bias vs variance compromise is managed independently in each block, allowing an enhancement if the energy of the unknown system is non-uniformly distributed. In order to give some insight on the behavior of the scheme, a theoretical analysis of the optimal scaling factors is developed. In addition, several sets of experiments are included to widely study the new scheme performance.
暂无评论