Robust Principal Component Analysis Research Articles

Unsupervised detecting anomalies in multivariate time series by Robust Convolutional LSTM Encoder–Decoder (RCLED)

Monitoring modern industrial systems generates a large amount of multivariate time series data. One of the critical tasks of monitoring these systems is anomaly detection. The auto-encoder has emerged as a promising solution for detecting anomalies in systems lacking explicit anomaly data in their historical records; however, its performance can be sensitive to noisy data. This work aims at developing a robust anomaly detection model based on the autoencoder for multivariate time series. The autoencoder architecture consists of convolution, long short-term memory, and self-attention layers for better extraction of complex features from multivariate time series. In addition, a novel training technique, inherited from the idea of robust principal component analysis, was developed to efficiently train the proposed model when the input data is affected by noise. To evaluate the model’s performance, we tested it on two data sets: synthetic, real-industrial, and three public data. We also compare the performance of our model to that of the other state-of-the-art models. The results show that the proposed model outperforms all the latest models, especially when the noise level is considerable.

Robust Principal Component Analysis via Joint Reconstruction and Projection.

Principal component analysis (PCA) is one of the most widely used unsupervised dimensionality reduction algorithms, but it is very sensitive to outliers because the squared l2 -norm is used as distance metric. Recently, many scholars have devoted themselves to solving this difficulty. They learn the projection matrix from minimum reconstruction error or maximum projection variance as the starting point, which leads them to ignore a serious problem, that is, the original PCA learns the projection matrix by minimizing the reconstruction error and maximizing the projection variance simultaneously, but they only consider one of them, which imposes various limitations on the performance of model. To solve this problem, we propose a novel robust principal component analysis via joint reconstruction and projection, namely, RPCA-RP, which combines reconstruction error and projection variance to fully mine the potential information of data. Furthermore, we carefully design a discrete weight for model to implicitly distinguish between normal data and outliers, so as to easily remove outliers and improve the robustness of method. In addition, we also unexpectedly discovered that our method has anomaly detection capabilities. Subsequently, an effective iterative algorithm is explored to solve this problem and perform related theoretical analysis. Extensive experimental results on several real-world datasets and RGB large-scale dataset demonstrate the superiority of our method.

Multidimensional Data Processing With Bayesian Inference via Structural Block Decomposition.

How to handle large multidimensional datasets, such as hyperspectral images and video information, efficiently and effectively plays a critical role in big-data processing. The characteristics of low-rank tensor decomposition in recent years demonstrate the essentials in describing the tensor rank, which often leads to promising approaches. However, most current tensor decomposition models consider the rank-1 component simply to be the vector outer product, which may not fully capture the correlated spatial information effectively for large-scale and high-order multidimensional datasets. In this article, we develop a new novel tensor decomposition model by extending it to the matrix outer product or called Bhattacharya-Mesner product, to form an effective dataset decomposition. The fundamental idea is to decompose tensors structurally in a compact manner as much as possible while retaining data spatial characteristics in a tractable way. By incorporating the framework of the Bayesian inference, a new tensor decomposition model on the subtle matrix unfolding outer product is established for both tensor completion and robust principal component analysis problems, including hyperspectral image completion and denoising, traffic data imputation, and video background subtraction. Numerical experiments on real-world datasets demonstrate the highly desirable effectiveness of the proposed approach.

Enhancing NILM classification via robust principal component analysis dimension reduction

Non-intrusive load monitoring (NILM) techniques estimate the consumption of individual appliances in a household or facility, based on aggregated reading from a centralized meter. Usually, NILM techniques are shown to be improved when various power features and additional power quality parameters are included. However, adding power features leads to increased time complexity which is a disadvantage to real-time operation. Previous attempt to operate a principal component analysis (PCA) method to reduce the dimension of the problem managed to improve the run time but with considerably low accuracy. To this end, we utilize a robust PCA approach, to mitigate the influence of outliers in the data as a measure for improved performance. The proposed procedure achieves extraordinary results with accuracy over 96% for 600 hours long record of power quality measurements of the consumption of seven appliances from the standard AMPds dataset.

Cybersecurity Challenges in Smart Grids: A Focus on Information Technology

The joining of information technology into smart grids has altered the energy area, improving proficiency and maintainability. Notwithstanding, this combination additionally delivers critical cybersecurity challenges. This paper digs into the investigation of cybersecurity challenges in smart grids, especially underlining the job of information technology. One basic viewpoint analyzed in this study is the use of organized Robust Principal Component Analysis (RPCA) with the Proximal Point identifier. Through definite estimations and examinations, the paper presents a thorough outline of the RPCA-based approach's viability. It gives experiences into the computational prerequisites for carrying out this method, featuring its true capacity in identifying oddities inside smart lattice frameworks. The exploration uses genuine data from the IEEE 30 and IEEE 118 power frameworks to assess the exhibition of the RPCA-based proximal tendency locater. Results exhibit promising results, including high detection probability and diminished recognizable proof latency. In addition, the review exhibits the calculation's ability to recognize False Data Injection Attacks (FDIA) with a great ID probability surpassing 95%. Besides, trial re-enactments led for both arbitrary and assigned assault situations on the IEEE 30 and IEEE 118 power frameworks display essentially lower detection latencies. These discoveries highlight the significance and viability of utilizing RPCA-based approaches in moderating cyber security dangers inside smart framework foundations.

Tensor robust principal component analysis via dual lp quasi-norm sparse constraints

As a pioneering method, tensor robust principal component analysis (TRPCA) can separate an underlying low-rank component and a sparse component from the original data by minimizing a convex objective function composed of tensor nuclear norm and l1-norm. However, it has two limitations. One is that tensor nuclear norm, as a constraint on the low-rank component, treats all singular values uniformly, ignoring the differences among singular values. In essence, this constraint is a sparse constraint that is imposed on singular values of the low-rank component by l1-norm. The other is that l1-norm is used as a constraint for the sparse component. However, l1-norm is a loose constraint, leading to the solutions of TRPCA deviating from the authentic ones. To alleviate these issues, we propose a TPRCA model called p-TRPCA that utilizes the lp-norm to impose sparse constraints on both the singular values and the sparse component simultaneously. The lp quasi-norm (0<p<1) is a tighter constraint than l1-norm, which enhances the low-rankness and sparsity of the proposed model. To solve p-TRPCA, we present an effective algorithm based on the alternating direction method of multipliers and also analyze its convergence. Extensive experiments are performed on data simulation, image recovery, and background modeling. Experimental results show that our p-TRPCA outperforms TRPCA and its variants.

Intelligent jamming detection‐based robust principal components analysis in communication system for security and defense

AbstractThreats to physical layer security from jamming attacks make wireless cognitive communication systems vulnerable. Global Positioning System signal is vulnerable to these attacks. Over the last decade, several types of jamming detection techniques have been proposed, antijamming‐based classical and machine learning (ML) techniques. Most of these techniques are inefficient in detecting jammers. Thus, there is a great need for efficient and quickest jamming detection technique‐based classifier using receiver operating characteristic (ROC) curve for different threshold values with high accuracy. In this work, we compare the efficiency of the proposed orthogonal distance (OD) and score distance (SD) method‐based robust principal component analysis (PCA) in ML classification in detecting jamming signals. Two hypotheses are proposed to distinguish between the presence and absence attack problem. Using this compressed data matrix obtained from modulated wideband converter (MWC) structure via centralized cooperation directly as input of the proposed classifier combined‐based ROC curve for real‐time detection scenarios. The performance of this proposed algorithm‐based robust PCA was evaluated and compared using the detection anomaly rate (DAR%), and false alarm rate (FAR%), area under curve (AUC), and accuracy. The performance of obtained results is good.

Tensor structure-based ground clutter suppression approaches for pulse doppler radar

ABSTRACT For pulse doppler (PD) radar, two tensor structure-based methods are proposed to improve the clutter suppression performance. One is tensor singular value decomposition (TSVD) and the other is tensor robust principal component analysis (TRPCA). Two algorithms both stack the range-Doppler matrices of multiple coherent processing intervals (CPI) into a tensor structure, which helps make the most of the temporal information across CPIs. TSVD utilizes the strong temporal correlation of ground clutter data to estimate the clutter subspace. Meanwhile, TRPCA exactly separates both low-rank clutter component and sparse target components from the time series range-Doppler tensor structure. Experimental results indicate that compared with conventional matrix-based schemes such as singular value decomposition (SVD) or singular value decomposition (RPCA), the proposed methods have significant advantages in keeping the target echo while suppressing the clutter. In addition, the computational complexity of each algorithm is analysed in detail.

Flat minima generalize for low-rank matrix recovery

Abstract Empirical evidence suggests that for a variety of overparameterized nonlinear models, most notably in neural network training, the growth of the loss around a minimizer strongly impacts its performance. Flat minima—those around which the loss grows slowly—appear to generalize well. This work takes a step towards understanding this phenomenon by focusing on the simplest class of overparameterized nonlinear models: those arising in low-rank matrix recovery. We analyse overparameterized matrix and bilinear sensing, robust principal component analysis, covariance matrix estimation and single hidden layer neural networks with quadratic activation functions. In all cases, we show that flat minima, measured by the trace of the Hessian, exactly recover the ground truth under standard statistical assumptions. For matrix completion, we establish weak recovery, although empirical evidence suggests exact recovery holds here as well. We complete the paper with synthetic experiments that illustrate our findings.

A multisensor approach coupled with multivariate statistics and geostatistics for assessing the status of land degradation: The case of soils contaminated in a former outdoor shooting range

Soil contamination in outdoor shooting ranges (OSRs) is a major threat for human health, particularly when, after the end of activities, the land is used for recreational areas or agricultural production. The status of land degradation of an OSR in southern Italy was assessed using a multisensor approach. It was based on: i) proximal sensors, including electromagnetic induction (EMI) for measuring soil electrical conductivity (ECa) and magnetic susceptibility (MSa), γ-ray spectrometry for K, eU and eTh analyses and ultrasonic penetrometry detecting cone index (CI) data representative of soil's strength, ii) field surveys on soil thickness (ST), and iii) laboratory analyses of potentially-toxic-elements (PTEs) by portable X-ray fluorescence spectrometry and polycyclic aromatic hydrocarbons (PAHs) by gas-chromatography. Spatial variability of measurements was modelled and mapped using geostatistical methods. The most densely measured covariate (i.e., the ECa of the topsoil) was used within kriging with external drift to improve the PTEs predictions. The PTEs maps were complemented by maps of spatial uncertainty. A robust multivariate principal component analysis (rPCA) was applied to proximal sensor and laboratory data and allowed to identify associations of PAHs, lead, CI with the topsoil ECa along the first component (PC1), highlighting the correlation between land anthropogenic effects and EMI measures; while the association between the ST (estimating the depth of underground travertine hard-layers) and the bottom soil ECa and MSa along the second component (PC2) evidenced the influence of soil stratigraphy on the EMI measures. This study demonstrates that the simultaneous use of different proximal sensors associated with laboratory analysis can allow to assess and model the spatial variability of the land degradation status of an OSR, including soil compaction, organic and inorganic contamination. The correlation between EMI data with the PTEs content highlights the potential of this technique in the field of soil contamination.

An Application of Robust Principal Component Analysis Methods for Anomaly Detection

Ensuring a secure network environment is crucial, especially with the increasing number of threats and attacks on digital systems. Implementing effective security measures, such as anomaly detection can help detect any abnormal traffic patterns. Several statistical and machine learning approaches are used to detect network anomalies including robust statistical methods. Robust methods can help identify abnormal traffic patterns and distinguish them from normal traffic accurately. In this study, a robust Principal Component Analysis (PCA) method called ROBPCA which is known for its extensive use in the literature of chemometrics and genetics is utilized for detecting network anomalies and compared with another robust PCA method called PCAGRID. The anomaly detection performances of these methods are evaluated by injecting synthetic traffic volume into a well-known traffic matrix. According to the application results, when the normal subspace is contaminated with large anomalies the ROBPCA method provides much better performance in detecting anomalies.

Principal components analysis and K-means clustering of till geochemical data: Mapping and targeting of prospective areas for lithium exploration in Västernorrland Region, Sweden

To achieve the demand for elements used for the green transition energy, such as lithium, it is necessary to recognize the spatial distribution of the concentrations of these elements in different earth materials such as bedrock and soil and to identify areas with anomalous concentrations of such elements (i.e., mineralization) for further exploration and hopefully exploitation. This study carried out multivariate statistical analyses on compositional (i.e., element concentration) data from till samples to recognize areas that likely contain lithium pegmatite mineralization in the Västernorrland region, central Sweden. We applied principal components analysis (PCA) and K-means clustering techniques to reveal regional-scale patterns in the till geochemical data. We demonstrate that these two methods have potential for recognition of geochemical anomalies related to the underlying bedrock geology as well as to mineralization. The results of PCA- and K-means clustering were validated using known occurrences of lithium mineralization. Two different datasets were compared; one containing all available geochemical data and the second containing only available trace elements in the dataset and it was found that anomalous clusters of samples defined by K-means clustering have anomalous multi-element signatures defined by robust PCA. This demonstrated that principal components are the continuous solutions to the discrete cluster members for the K-means clustering. The results show that both PCA and K-means clustering of till geochemical datasets at the early stages of exploration and target generation may reveal useful information that can be used to identify potential areas for more detailed mapping or exploration activities.

Tensor robust PCA with nonconvex and nonlocal regularization

Tensor robust principal component analysis (TRPCA) is a classical way for low-rank tensor recovery, which minimizes the convex surrogate of tensor rank by shrinking each tensor singular value equally. However, for real-world visual data, large singular values represent more significant information than small singular values. In this paper, we propose a nonconvex TRPCA (N-TRPCA) model based on the tensor adjustable logarithmic norm. Unlike TRPCA, our N-TRPCA can adaptively shrink small singular values more and shrink large singular values less. In addition, TRPCA assumes that the whole data tensor is of low rank. This assumption is hardly satisfied in practice for natural visual data, restricting the capability of TRPCA to recover the edges and texture details from noisy images and videos. To this end, we integrate nonlocal self-similarity into N-TRPCA, and further develop a nonconvex and nonlocal TRPCA (NN-TRPCA) model. Specifically, similar nonlocal patches are grouped as a tensor and then each group tensor is recovered by our N-TRPCA. Since the patches in one group are highly correlated, all group tensors have strong low-rank property, leading to an improvement of recovery performance. Experimental results demonstrate that the proposed NN-TRPCA outperforms existing TRPCA methods in visual data recovery. The demo code is available at https://github.com/qguo2010/NN-TRPCA.

Superposed Atomic Representation for Robust High-Dimensional Data Recovery of Multiple Low-Dimensional Structures

This paper proposes a unified Superposed Atomic Representation (SAR) framework for high-dimensional data recovery with multiple low-dimensional structures. The data can be in various forms ranging from vectors to tensors. The goal of SAR is to recover different components from their sum, where each component has a low-dimensional structure, such as sparsity, low-rankness or be lying a low-dimensional subspace. Examples of SAR include, but not limited to, Robust Sparse Representation (RSR), Robust Principal Component Analysis (RPCA), Tensor RPCA (TRPCA), and Outlier Pursuit (OP). We establish the theoretical guarantee for SAR. To further improve SAR, we also develop a Weighted SAR (WSAR) framework by paying more attention and penalizing less on significant atoms of each component. An effective optimization algorithm is devised for WSAR and the convergence of the algorithm is rigorously proved. By leveraging WSAR as a general platform, several new methods are proposed for high-dimensional data recovery. The experiments on real data demonstrate the superiority of WSAR for various data recovery problems.

Abstract 4954: Improving microbial detection in cancer tissue samples with computational host depletion using the pangenome

Abstract Sensitive removal of host genetic information (i.e., host depletion) is a key first step towards accurate classification of potential microbial DNA sequences in low-biomass samples, such as human tumors. Inadequate filtration of human reads prior to microbial classification can cause false positives, biasing biological conclusions. Previous computational host depletion pipelines predominantly employed the human reference genome GRCh38 (hg38), which lacks genetic diversity of the human population and excludes substantial portions of the Y-chromosome. To address these deficiencies, we constructed a pipeline to maximally remove human reads that incorporates hg38, the first complete genome build T2T-CHM13v2.0 (T2T), and the 47 diploid genomes from the Human Pangenome Reference Consortium (HPRC) [1]. Reads passing maximal host depletion and quality control are considered sufficiently cleaned for downstream microbial analyses. We first evaluated the specificity of our pipeline by simulating Illumina whole genome sequencing (WGS) data from 10 heldout HPRC humans, whose genomes were excluded from the host depletion pipeline. We found significant improvement in the median number of human reads remaining across successive host depletion methods (hg38, 0.05369%; hg38+T2T, 0.00535%; hg38+T2T+HPRC, 0.00093%). For sensitivity, we applied our pipeline to simulated data from 962 microbes in the FDA-ARGOS dataset and found minimal removal of true microbial reads (median 99.9390% reads retained). As a real world example, we evaluated our pipeline on 583 metastatic colorectal cancer tissue (WGS) samples from the Hartwig Medical Foundation (HMF) database (median 106X coverage)[2] and found successive decreases in median number of reads retained (hg38, 185,135; hg38+T2T, 92,951; hg38+T2T+HPRC, 10,314). Using Woltka for taxonomic classification against database RefSeq210, we assessed biological sex similarity across the HMF samples using Robust Aitchison PCA (RPCA). Notably, microbial data following hg38-only host depletion showed significant microbial differences between male and female-labeled samples (p=0.00001), which was biologically unexpected. However, our updated pipeline (hg38+T2T+HPRC) removed the false sex-based separation in the microbial data (p=0.1837). Thus, utilization of diverse sets of complete human genomes for computational host depletion can mitigate artifactual bias, enabling more sensitive and specific host-microbe conclusions in cancer research. [1] Liao, et al. Nature 617, 312-324 (2023). [2] Priestley, et al. Nature 575, 210-216 (2019). Citation Format: Caitlin Guccione, Lucas Patel, Daniel McDonald, Cameron Martino, Antonio González, Gregory D. Sepich-Poore, Rob Knight, Kit Curtius. Improving microbial detection in cancer tissue samples with computational host depletion using the pangenome [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4954.

Identifying high potential gold mineralization using geological and stream sediment geochemical data: A case study from western Dangreyongcuo area in the central Lhasa terrane, Tibet, China

Northern Tibet exhibits abundant placer gold deposits, with a longstanding history of mining activities. Despite the widespread placer gold occurrences, the reports on the primary rock gold deposits or occurrences remain scarce. The central Lhasa terrane in northern Tibet hosts well-developed Zenong Group volcanic rocks, indicating great potential for the exploration of porphyry-epithermal copper (gold) deposits. Here, we select the Dangreyongcuo area in this region to investigate its metallogenetic potentiality, through a comprehensive 1:50,000 stream sediment geochemical survey, analyzing 12 elements (Au, As, Sb, Ag, Te, Cu, Pb, Zn, W, Sn, Bi, and Mo), and a detailed geological mapping. We aim to identify gold-related mineralization anomalies and provide clues for further bedrock gold exploration in this region. Based on the stream sediment geochemical data, zones 1–4, have been delineated using single-element indicator Au and element association indicator Au + As + Sb, which was determined through robust principal component analysis (RPCA) and using the exploratory data analysis (EDA) method. Subsequent validation exercises evaluating these four delineated zones reveal that Zone 1, Xinlong-Langmeila area, is the most promising prospect due to its high element concentrations, large size, and marked coincidence of element assemblages (also referred to as high-large-coincidence anomalies; HLCAs) and its characteristics are in line with those of epithermal deposits. Notably, the Xinlong deposit represents the first documented epithermal gold deposit hosted within the Zenong Group volcanic rocks. The identification of this deposit offers valuable insights and strategic directions for the broader understanding of Zenong Group volcanic rocks across the central Lhasa terrane. Additionally, subsequent strategies have been proposed for mineral exploration at both Xinlong camp and regional scales.

RPCA-based techniques for pattern extraction, hotspot identification and signal correction using data from a dense network of low-cost NO2 sensors in London

High-density low-cost air quality sensor networks are a promising technology to monitor air quality at high temporal and spatial resolution. However the collected data is high-dimensional and it is not always clear how to best leverage this information, particularly given the lower data quality coming from the sensors. Here we report on the use of robust Principal Component Analysis (RPCA) using nitrogen dioxide data obtained from a recently deployed dense network of 225 air pollution monitoring nodes based on low-cost sensors in the Borough of Camden in London. RPCA addresses the brittleness of singular value decomposition towards outliers by using a decomposition of the data into low-rank and sparse contributions, with the latter containing outliers. The modal decomposition enabled by RPCA identifies major periodic patterns including spatial and temporal bias, dominant spatial variance, and north-south bias. The five most descriptive components capture 98 % of the data's variance, achieving a compression by a factor of 1500. We present a new technique that uses the sparse part of the data to identify hotspots. The data indicates that at the locations of the top 15 % most susceptible nodes in the network, the model identifies 23 % more hotspots than in all other locations combined. Moreover, the median hotspot event at these at-risk locations exceeds the mean NO2concentration by 33μg/m3. We show the potential of RPCA for signal correction; it corrects random errors yielding a reference signal with R2>0.8. Moreover, RPCA successfully reconstructs missing data from a sensor with R2=0.72 from the rest of the sensor network, an improvement upon PCA of around 50 %, allowing air quality estimations even if a sensor is out of use temporarily.

Traceability of Microplastic Fragments from Waste Plastic Express Packages Using Near-Infrared Spectroscopy Combined with Chemometrics.

The pollution from waste plastic express packages (WPEPs), especially microplastic (MP) fragments, caused by the blowout development of the express delivery industry has attracted widespread attention. On account of the variety of additives, strong complexity, and high diversity of plastic express packages (PEPs), the multi-class classification of WPEPs is a typical large-class-number classification (LCNC). The traceability and identification of microplastic fragments from WPEPs is very challenging. An effective chemometric method for large-class-number classification would be very beneficial for the comprehensive treatment of WPEP pollution through the recycling and reuse of waste plastic express packages, including microplastic fragments and plastic debris. Rather than using the traditional one-against-one (OAO) and one-against-all (OAA) dichotomies, an exhaustive and parallel half-against-half (EPHAH) decomposition, which overcomes the defects of the OAO's classifier learning limitations and the OAA's data proportion imbalance, is proposed for feature selection. EPHAH analysis, combined with partial least squares discriminant analysis (PLS-DA) for large-class-number classification, was performed on 750 microplastic fragments of polyethylene WPEPs from 10 major courier companies using near-infrared (NIR) spectroscopy. After the removal of abnormal samples through robust principal component analysis (RPCA), the root mean square error of cross-validation (RMSECV) value for the model was reduced to 0.01, which was 21.5% lower than that including the abnormal samples. The best models of PLS-DA were obtained using SNV combined with SG-17 smoothing and 2D (SNV+SG-17+2D); the latent variables (LVs), the error rates of Monte Carlo cross-validation (ERMCCVs), and the final classification accuracies were 6.35, 0.155, and 88.67% for OAO-PLSDA; 5.37, 0.103, and 87.33% for OAA-PLSDA; and 3.12, 0.054, and 96.00% for EPHAH-PLSDA. The results showed that the EPHAH strategy can completely learn the complex LCNC decision boundaries for 10 classes, effectively break the tie problem, and greatly improve the voting resolution, thereby demonstrating significant superiority to both the OAO and OAA strategies in terms of classification accuracy. Meanwhile, PLS-DA further maximized the covariance and data interpretation abilities between the potential variables and categories of microplastic debris, thereby establishing an ideal performance identification model with a recognition rate of 96.00%.

A Clutter Removal Method Based on the F-K Domain for Ground-Penetrating Radar in Complex Scenarios

Ground-penetrating radar (GPR) is a classic geophysical exploration method that utilizes the emission and reception of electromagnetic waves to non-destructively detect target objects in the target medium. It has been widely applied in various fields such as pipeline detection, cavity detection, and rebar detection. However, GPR systems are susceptible to environmental clutter interference, which poses challenges for data interpretation and subsequent processing. In this paper, the separability of clutter and target signal in the frequency-wavenumber (F-K) domain is validated through modeling, leading to the proposal of a comprehensive clutter removal method based on the F-K domain for complex scenarios. The direct coupling wave is initially eliminated by applying a peak matching mean subtraction filter, which avoids the artifacts. Subsequently, the F-K domain transformation is performed and surface clutter undulations are effectively removed using a method based on singular value decomposition and k-means clustering. Finally, an angle filter with Gaussian tapering at the edges is designed based on physical models to efficiently eliminate linear interference without undesired ringing interference. The commonly used clutter removal algorithms, including mean subtraction (MS), singular value decomposition (SVD), robust principal component analysis (RPCA), and traditional F-K filtering methods, are compared with the proposed algorithm on both the numerical simulated data and actual GPR data. The results from visual and quantitative analysis confirm that our proposed method is more effective than current commonly used clutter suppression algorithms. We have successfully enhanced the Signal-to-Clutter Ratio (SCR) of the GPR data, resulting in an Improvement Factor (IF) of 30.63 dB, 23.59 dB, and 30.60 dB for simulated data, experimental data, and TU1208 public data, respectively. The detection capability of buried targets is enhanced, thereby establishing a solid foundation for subsequent data interpretation and target identification.

Lithium exploration targeting through robust variable selection and deep anomaly detection: An integrated application of sparse principal component analysis and stacked autoencoders

Lithium is a strategic metal for high-technology industries that plays a vital role in realizing electromobility and effective energy storage for smartphones and electric/hybrid vehicles, in addition to being important in accumulating energy from renewable sources. Motivated by this, the development of state-of-the-art workflows is imperative to make Li exploration more efficient and guarantee a sustainable supply to the demanding markets in the future. In this paper, a computational protocol was described to address two critical challenges raised in the Li exploration: (i) selection of mineralization-related pathfinder variables and (ii) recognition of multi-element geochemical anomalies related to ore-forming processes. Robust sparse principal component analysis (SPCA) was employed to reduce active (non-zero) loading coefficients and straightforward selection of Li pathfinder variables. The selected variables were then fed into the training network of stacked autoencoders (SAE) to learn deep representations of multivariate input signals and quantify reconstruction errors linked to Li-vectoring geochemical anomalies. As a representative example, the proposed workflow, i.e., SPCA + SAE, was implemented in the R programming language and applied to a real-case experiment with stream sediment geochemical data pertaining to the Moalleman district, NE Iran. Moreover, compositional robust principal component analysis (RPCA) and Mahalanobis distance (MD) technique were adopted to constitute two comparative models, RPCA + SAE and SPCA + MD, as benchmarks to judge the competence of the SPCA + SAE model in variable selection and anomaly detection, respectively. A performance appraisal by success-rate curves and relevant area under the curves (AUCs) indicated that the SPCA + SAE model brings, by calculating the highest AUC, spatial patterns that are more promising for vectoring towards Li-mineralized grounds. Moreover, AUCSPCA+MD was measured to be greater than AUCRPCA+SAE, suggesting robust variable selection is a more important paradigm than deep anomaly detection for Li exploration targeting. Student's t–statistic method was eventually applied to the anomaly map from the SPCA + SAE model to define relevant thresholds for narrowing down prospective areas for the next round of Li exploration within the study area.