GA Based Information Retrieval System Using Dimensionality Reduction Techniques

In this study, principal component analysis (PCA), singular value decomposition (SVD) and independent component analysis (ICA), which are statistical signal processing methods for clutter reduction, are explained in ultra wideband (UWB) radars. In addition, how to use these methods in through wall imaging (TWI) and time/performance relationship on ARM Cortex A-8 processors which have floating-point mathematical operation support is examined by using OpenCV library. In the conclusion part, the results obtained from a metal behind a wall was made of ytong bricks is discussed.

Ambient fluctuations and the generation of coherent convective structures are investigated as a function of vertical to toroidal magnetic field ratio (Bz/Bt) in the spherical tokamak QUEST [K. Hanada et al., Plasma Fusion Research 5, S1007]. Fluctuation signals are obtained at a high spatial and temporal resolution using the fast visible camera. The coherent structures (blobs) were seen to be apparent only beyond a critical Bz/Bt ratio (~ 10). This clearly indicates a shift from the drift to the interchange mode with the increase in vertical magnetic field. The blobs are propagated radially at a high speed ~ 10 m/s which is ~ 1/10 of the ion acoustic speed in QUEST. Lifetime of the blobs is about 700 μs. Second part of the study comprises of the investigation of the finer structures in the current ramp-up phase which is characterized by lower levels of fluctuations. Singular value decomposition (SVD) and principal component analysis (PCA) techniques are applied on the fast camera images to reveal the structures which are otherwise embedded in the overall intensity image. Electron banana orbits were seen to form from the slab annular plasma and grow wider to eventually shape the plasma current front.

A method is defined and tested for the characterization of agricultural land from multi-spectral imagery, based on singular value decomposition (SVD) and key vector analysis. The SVD technique, which bears a close resemblance to multivariate statistic techniques, has previously been successfully applied to problems of signal extraction for marine data and forestry species classification. In this study the SVD technique is used as a classifier for agricultural regions, using airborne Daedalus ATM data, with 1 m resolution. The specific region chosen is an experimental research farm in Bavaria, Germany. This farm has a large number of crops, within a very small region and hence is not amenable to existing techniques. There are a number of other significant factors which render existing techniques such as the maximum likelihood algorithm less suitable for this area. These include a very dynamic terrain and tessellated pattern soil differences, which together cause large variations in the growth characteristics of the crops. The SVD technique is applied to this data set using a multi-stage classification approach, removing unwanted land-cover classes one step at a time. Typical classification accuracy's for SVD are of the order of 85-100%. Preliminary results indicate that it is a fast and efficient classifier with the ability to differentiate between crop types such as wheat, rye, potatoes and clover. The results of characterizing 3 sub-classes of Winter Wheat are also shown.

Retinopathy based techniques are preferred for diabetes detection due to their advantage of bloodless diagnosis. In concert to similar works currently going on, we propose, such a diabetes detection method using Artificial Neural Network (ANN) and a feature set formed by adopting Singular Value Decomposition (SVD) and Principal Component Analysis (PCA). A robust and computationally efficient approach for the localization of the different features in a fundus retinal image is presented in this work. The work proposes certain approaches for extraction of diabetes retinopathy features using SVD and PCA and applying the composite form to ANN for training. The detection of hemorrhages and exudates are important in order to diagnose diabetes retinopathy for preventing loss of eye sight. Experimental results show that the ANN-SVD+PNN composition is a reliable means of diabetes detection with less computational complexity and high accuracy.

DeepTensor is a computationally efficient framework for low-rank decomposition of matrices and tensors using deep generative networks. We decompose a tensor as the product of low-rank tensor factors (e.g., a matrix as the outer product of two vectors), where each low-rank tensor is generated by a deep network (DN) that is trained in a self-supervised manner to minimize the mean-square approximation error. Our key observation is that the implicit regularization inherent in DNs enables them to capture nonlinear signal structures (e.g., manifolds) that are out of the reach of classical linear methods like the singular value decomposition (SVD) and principal components analysis (PCA). Furthermore, in contrast to the SVD and PCA, whose performance deteriorates when the tensor's entries deviate from additive white Gaussian noise, we demonstrate that the performance of DeepTensor is robust to a wide range of distributions. We validate that DeepTensor is a robust and computationally efficient drop-in replacement for the SVD, PCA, nonnegative matrix factorization (NMF), and similar decompositions by exploring a range of real-world applications, including hyperspectral image denoising, 3D MRI tomography, and image classification. In particular, DeepTensor offers a 6 dB signal-to-noise ratio improvement over standard denoising methods for signal corrupted by Poisson noise and learns to decompose 3D tensors 60 times faster than a single DN equipped with 3D convolutions.

Low-rank prior-based Fast-RPCA for clutter filtering and noise suppression in non-contrast ultrasound microvascular imaging

Aiming at the problem of human behavior recognition, this paper proposes a human behavior classification algorithm based on radial basis function (BP) neural network. The singular value decomposition (SVD) and principal component analysis (PCA) are used for feature extraction, and the linear discriminant analysis (LDA) is used for feature transformation to make it easier to distinguish between various categories. Finally, BP neural network classifier is used to recognize human behavior. The experimental results show that compared with the nearest neighbor classification and k-nearest neighbor classification (kNN), the proposed algorithm has a higher recognition rate. The result of BP neural network classifier is the best, it can fit the boundary of each category well.

Principal component analysis (PCA), which is designed to look at internal modes of variability, has often been applied beyond its intended design to study coupled modes of variability in combined datasets, also referred to as combined PCA. There are statistical techniques better suited for this purpose such as singular value decomposition (SVD) and canonical correlation analysis (CCA). In this paper, a different technique is examined that has not often been applied in climate science, that is, redundancy analysis (RA). Similar to multivariate regression, RA seeks to maximize the variance accounted for in one random vector that is linearly regressed against another random vector. RA can be used for forecasting and prediction studies of the climate system. This technique has the added advantage that the time-lagged redundancy index offers a robust method of identifying lead–lag relations among climate variables. In this study, combined PCA and RA of global sea surface temperatures (SSTs) and sea level pressures (SLPs) are carried out for the National Centers for Environmental Prediction (NCEP) reanalysis data and a simulation of the Canadian Centre for Climate Modeling and Analysis (CCCma) climate model. A simplified state-space model is also constructed to aid in the diagnosis and interpretation of the results. The relative advantages and disadvantages of combined PCA and RA are discussed. Overall, RA tends to provide a clearer and more consistent picture of the underlying physical processes than combined PCA.

Cells shape or density is an important marker of tissues pathology. However, individual cells are difficult to observe in thick tissues frequently presenting highly scattering structures such as collagen fibers. Endogenous techniques struggle to image cells in these conditions. Moreover, exogenous contrast agents like dyes, fluorophores or nanoparticles cannot always be used, especially if non-invasive imaging is required. Scatterers motion happening down to the millisecond scale, much faster than the still and highly scattering structures (global motion of the tissue), allowed us to develop a new approach based on the time dependence of the FF-OCT signals. This method reveals hidden cells after a spatiotemporal analysis based on singular value decomposition and wavelet analysis concepts. It does also give us access to local dynamics of imaged scatterers. This dynamic information is linked with the local metabolic activity that drives these scatterers. Our technique can explore subcellular scales with micrometric resolution and dynamics ranging from the millisecond to seconds. By this mean we studied a wide range of tissues, animal and human in both normal and pathological conditions (cancer, ischemia, osmotic shock…) in different organs such as liver, kidney, and brain among others. Different cells, undetectable with FF-OCT, were identified (erythrocytes, hepatocytes…). Different scatterers clusters express different characteristic times and thus can be related to different mechanisms that we identify with metabolic functions. We are confident that the D-FFOCT, by accessing to a new spatiotemporal metabolic contrast, will be a leading technique on tissue imaging and for better medical diagnosis.

The role of subsurface temperature variability in modulating El Niño–Southern Oscillation (ENSO) properties is examined using an intermediate coupled model (ICM), consisting of an intermediate dynamic ocean model and a sea surface temperature (SST) anomaly model. An empirical procedure is used to parameterize the temperature of subsurface water entrained into the mixed layer (Te) from sea level (SL) anomalies via a singular value decomposition (SVD) analysis for use in simulating sea surface temperature anomalies (SSTAs). The ocean model is coupled to a statistical atmospheric model that estimates wind stress anomalies also from an SVD analysis. Using the empirical Te models constructed from two subperiods, 1963–79 (T63–79e) and 1980–96 (T80–96e), the coupled system exhibits strikingly different properties of interannual variability (the oscillation period, spatial structure, and temporal evolution). For the T63–79e model, the system features a 2-yr oscillation and westward propagation of SSTAs on the equator, while for the T80–96e model, it is characterized by a 5-yr oscillation and eastward propagation. These changes in ENSO properties are consistent with the behavior shift of El Niño observed in the late 1970s. Heat budget analyses further demonstrate a controlling role played by the vertical advection of subsurface temperature anomalies in determining the ENSO properties.

Surrogate modeling is essential in reducing computational costs for history-matching applications. Yet, traditional deep learning-based surrogate models cannot cope with high dimensional input parameters, such as the permeability field. This work introduces a robust method to automate the history matching process utilizing the Bayesian inversion assisted by a hybrid convolutional neural network and long short-term memory (CNN-LSTM) model and principal component analysis (PCA) method. The method includes five main steps. Step 1: Generate a high-spatial permeability field using a geostatistical approach. Step 2: use the PCA to reduce the dimensionality of the permeability fields, followed by using PCA to generate permeability fields and perform simulations. Step 3: construct the CNNLSTM to map the nonlinear relationship between the extracted features from PCA and the sequential outputs, such as the pressure response. Here, Bayesian optimization is employed to automate hyperparameter tuning. Step 4: perform the Bayesian inversion to inverse the high dimensional inputs, e.g., permeability field, in which the CNN-LSTM serves as the forward model to reduce the computational cost. The inversed PCA features are then fed into the PCA to recover the high dimensional inputs. Step 5: check convergence and if the errors are significant between the inversed high dimensional permeability field and the ground truth, revisit the construction of the CNN-BiLSTM and the prior information for the uncertainty parameters. A 2D reservoir model demonstrates the proposed history-matching method. We can inverse the high dimensional inputs (e.g., permeability field) with minor errors between the prediction and ground truth. We propose a Bayesian inversion assisted by a hybrid CNN-LSTM model and PCA method for high-dimensional parameter inversion, which is superior to the traditional models regarding accuracy and efficiency. This method enables us to perform history matching for reservoir simulation with high dimensional inputs and significant uncertainties.

Ultraviolet absorption and Raman spectra of p-nitroaniline (pNA) have been measured in acetonitrile/CCl4 mixed solvents for seventeen different fractions of acetonitrile. The observed spectral changes were interpreted with the use of a combined singular value decomposition (SVD) analysis, in which the ultraviolet absorption and Raman spectra were combined into one set of spectra. A least-squares fitting analysis was carried out for the SVD decomposed spectra and their amplitudes, using a model in which three distinct species of pNA, the 1 : 0, 1 : 1, and 1 : 2 species, exist in equilibrium. Here, the 1 : 0, 1 : 1, and 1 : 2 species have 0, 1, and 2 acetonitrile molecules associated with pNA. Three intrinsic ultraviolet absorption and Raman spectra have been obtained as a result of the fitting analyses. The three intrinsic spectra show different degrees of charge transfer in the three different solvated forms of pNA. Two equilibrium constants, K1 for the equilibrium between the 1 : 0 and 1 : 1 species and K2 for 1 : 1 and 1 : 2, have been determined. These equilibrium constants give the standard free energy changes for the association of acetonitrile with pNA. A microscopic and quantitative view on the solvation of pNA in acetonitrile/CCl4 mixed solvents has thus been established.

This article develops and compares health indices using different approaches namely singular value decomposition, average value of the cumulative feature and Mahalanobis distance for assessing the rolling element bearing condition. The vibration signals for four conditions of rolling element bearing are acquired from a customized bearing test rig under variable load conditions. Seventeen statistical features are extracted from wavelet coefficients of the denoised signals. Feature selection is performed using singular value decomposition and kernel Fisher discriminant analysis. These selected features are used in these three approaches to develop health indices. Finally, a comparison of the three proposed approaches is made to select the best approach which can be effectively used for fault diagnosis of rolling element bearings.

A singular value decomposition (SVD) analysis between the Northern Hemisphere winter-mean 500-hPa geopotential height (Z500) and near-global surface temperature (land-surface air temperature + sea surface temperature) fields was conducted for the recent three decades. The leading mode, referred to as Polar/Eurasia (PEA), is considered as a hemispheric-wide atmospheric internal variability rather than a local teleconnection pattern. Winter mean Z500 of the leading mode is closely related to Eurasian surface air temperature from autumn to spring throughout the period. Both interannual and interdecadal time scales are dominant during the three decades in the time scales. No linear relationship between the tropical SST anomalies is found in this mode. The second and third modes can be identified as the Pacific/North American (PNA) and Western Pacific (WP)-type teleconnection patterns, respectively. Although they are also considered as internal modes of atmospheric variabilities, their interannual variabilities are also influenced by the tropical Pacific SST anomalies. These results are compared with the more popular SVDanalysis between geopotential height and SST, for which the leading mode exhibits the so-called "ENSO (El Niño/Southern Oscillation) related" winter circulation pattern. It is shown that this ENSO-related mode can be reconstructed from the linear combination of the second (PNA) and third (WP) modes of the present SVD mentioned above. These results suggest that inter-El Niño differences of atmospheric circulation in midlatitudes can be interpreted as the different strengths of the PNA and WP patterns on one hand, and on the other hand by the different magnitude of the PEA, which is linearly independent of ENSO. The characteristics of the vertical structure of the atmospheric variability were further examined by conducting a SVD analysis of 100-hPa temperature and Z500. The PEA and PNA patterns in the Z500 field extracted as the first and second SVD modes, respectively, exhibit contrasting features in association with lower stratospheric temperature. The PNA pattern is confined to the troposphere, while PEA extends further into the stratosphere. This suggests they are produced by different mechanisms.

In radar detection, small and slow targets are easily overwhelmed by strong clutter. Traditional methods, such as singular value decomposition (SVD) and robust principal component analysis (RPCA), can suppress clutter and recover targets by using low-rank and sparse models. However, these methods rely on fixed prior information, which lacks adaptivity and suffers from unfavorable extensive manual hyperparameter tuning. To address these issues, a data-driven deep network model combined with an iterative algorithm called unfolding joint sparse recovery network (UFJSR-Net) is proposed to achieve the improved target detection performance. First, a joint sparse recovery (JSR) model is established and the fast iterative shrinkage/thresholding algorithm (FISTA) is derived to solve this model. Then, one iteration consisting of a linear operation and a nonlinear one can be recast into a single network layer and the stacking and combination of all layers will form the UFSJR-Net. Finally, the properties of the target and clutter can be learned by paired inputs and outputs training data to optimize the hyperparameters in the iterative algorithm to obtain the JSR model. Experiments on simulation data and measured radar data demonstrate that the proposed method exhibits advantages in detection performance over traditional decomposition methods under strong clutter with different intensities.