Gil Goldshlager, Jiang Hu, Lin Lin · ArXiv.org · 2025
Due to the ever growing amounts of data leveraged for machine learning and scientific computing, it is increasingly important to develop algorithms that sample only a small portion of the data at a time. In the case of linear least-squares, the randomized block Kaczmarz method (RBK) is an appealing example of such an algorithm, but its convergence is only understood under sampling distributions that require potentially prohibitively expensive preprocessing steps. To address this limitation, we analyze RBK when the data is sampled uniformly, showing that its iterates converge in a Monte Carlo sense to a $\textit{weighted}$ least-squares solution. Unfortunately, for general problems the bias of the weighted least-squares solution and the variance of the iterates can become arbitrarily large. We show that these quantities can be rigorously controlled by incorporating regularization into the RBK iterations, yielding the regularized algorithm ReBlocK. Numerical experiments including examples arising from natural gradient optimization demonstrate that ReBlocK can outperform both RBK and minibatch stochastic gradient descent for inconsistent problems with rapidly decaying singular values.
Cited by 0Open access
Hernandez, Hortencia Josefina · scholarworks - UTEP (The University of Texas at El Paso) · 2022
Graph theory modeling is a common modeling approach in neurobiology research studies. These models are useful since they describe patterns of connection for regions of interest in the brain using resting state fMRI images. The standard rule of thumb is to threshold the observed activation levels prior to model building. It is reasonable to assume that the use of this threshold affects the statistical distribution of commonly reported centrality metrics from the graph theory model, such as degree, betweenness, and closeness. In this study we examine the differential effect of using the standard approaches versus alternative direct thresholds and incorporation of thresholds through soft and hard covariance estimation. Along with the way it is viewed we care about the way we preprocess the data. Results indicate that direct thresholding is a more reliable preprocessing strategy, but soft thresholding of the covariance matrix may be a promising alternative in particular settings.
Cited by 0
G Ranganathan · Journal of Innovative Image Processing · 2021
In the near future, deep learning algorithms will be incorporated in several applications for assisting the human beings. The deep learning algorithms have the tendency to allow a computer to work on its assumption. Most of the deep learning algorithms mimic the human brain's neuron connection to leverage an artificial intelligence to the computer system. This helps to improve the operational speed and accuracy on several critical tasks. This paper projects the blocks, which are required for the incorporation of deep learning based algorithm. Also, the paper attempts to deeply analyze the necessity of the preprocessing step over several deep learning based applications.
Cited by 131Open access
D.C. Ross · IEEE Antennas and Propagation Magazine · 1993
All finite-element mesh-generation packages output a so-called universal file which, as a minimum, contains a list of the node coordinates, a table of the nodes which form each element, and a list of the materials filling the element. Algorithms that make use of this information to generate additional information or tables, which are typically required in any electromagnetic analysis code, are given. The given code fragments are in C to avoid explicit dimensioning of arrays and facilitate the execution of recursive routines.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Cited by 2
Sandra Castillo, Peddinti Gopalacharyulu, Laxman Yetukuri et al. · Chemometrics and Intelligent Laboratory Systems · 2011
Cited by 169
Njideka Chiamaka Okafor, Bill Cassidy, Claire O’Shea et al. · Lecture notes in computer science · 2024
Cited by 5
Sergei O. Kuznetsov, I.V. Nuidel, Andrey I. Panfilov et al. · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1998
The problem of image pre-processing by neuron-like algorithms concerns development of systems and methods of image processing in parallel regime. The processing algorithms of grey-tone images to sets of simplified binary images by software of neuron-like filtration (based on the models of homogeneous neural networks) are considered in the paper and partially realized. The time of image processing is calculated. Carrying out of the required set of operations with image is possible according to the convolution function view. The neural network parameters (the type of coupling function and type of element response to the external influence) are selected in accordance with the solution which we want to receive. The program uses function similar to the Mask Convolution function. This function as any other functions basic for image processing are realized in the Intel Corporation `Recognition Primitives Library (RPL) for the Pentium Processor' with great efficiency. Library works under Windows NT. Performed calculations confirm that neuron-like model may be applied for real time image processing using RPL.
Cited by 2
Colin White · Lecture notes in computer science · 2015
Cited by 5
Iris Kameny, H. Barry Ritea · 1970
Abstract : The document provides a detailed description and analysis of the preprocessing and segmentation procedures used in the Vicens-Reddy speech recognition system.
Cited by 0
S. Raguvir, D. Radha · Lecture notes in computational vision and biomechanics · 2019
Cited by 4
Pat Mcingvale, R.D. Guyton · 2005
Cited by 0
S. Raguvir, Dr.P. Radha, Durai Pandian et al. · 2019
Cited by 0
Abdul Quadir, Sanika Kulkarni, Christy Jackson Joshua et al. · Biomedicines · 2023
There has been a sharp increase in liver disease globally, and many people are dying without even knowing that they have it. As a result of its limited symptoms, it is extremely difficult to detect liver disease until the very last stage. In the event of early detection, patients can begin treatment earlier, thereby saving their lives. It has become increasingly popular to use ensemble learning algorithms since they perform better than traditional machine learning algorithms. In this context, this paper proposes a novel architecture based on ensemble learning and enhanced preprocessing to predict liver disease using the Indian Liver Patient Dataset (ILPD). Six ensemble learning algorithms are applied to the ILPD, and their results are compared to those obtained with existing studies. The proposed model uses several data preprocessing methods, such as data balancing, feature scaling, and feature selection, to improve the accuracy with appropriate imputations. Multivariate imputation is applied to fill in missing values. On skewed columns, log1p transformation was applied, along with standardization, min-max scaling, maximum absolute scaling, and robust scaling techniques. The selection of features is carried out based on several methods including univariate selection, feature importance, and correlation matrix. These enhanced preprocessed data are trained on Gradient boosting, XGBoost, Bagging, Random Forest, Extra Tree, and Stacking ensemble learning algorithms. The results of the six models were compared with each other, as well as with the models used in other research works. The proposed model using extra tree classifier and random forest, outperformed the other methods with the highest testing accuracy of 91.82% and 86.06%, respectively, portraying our method as a real-world solution for detecting liver disease.
Cited by 102Open access
Viktor Shynkarenko, O.V. Makarov · PROBLEMS IN PROGRAMMING · 2023
To verify the hypothesis about decrease in time of sorting by algorithms of different computational complexity experiments have been conducted. Several ideas on deterministic preprocessing of data arrays for sorting algorithms have been tested. The following algorithms are proposed: quick preprocessing – prediction of the index of an element in a sorted array and permutation, preprocessing with memory - prediction and permutation with memorization of previously set elements, preprocessing with reordering – reverting sequences of elements sorted in reverse order. Also proposed block variations of quick and preprocessing with memory, which are performed for parts of the array of a given length. It has been defined that the higher efficiency of preprocessing is achieved by using with sorting algorithms, which are significantly accelerated on sorted (or almost sorted) arrays of data. Block preprocessing methods can be performed faster due to the possibility of avoiding cache misses, but show a lower percentage of array sorting. Experiments were conducted to evaluate the effectiveness of various sorting algorithms after and together with the proposed preprocessing methods.
Cited by 2Open access
Ramón O. Guardado-Medina, Luis E. Mendoza-Luna, Jorge A. Ortiz-Ramirez et al. · OALib · 2014
This paper presents the performance analysis of preprocessing algorithms for enhancement features on mammograms with the objective of improving future clustering processing investigations, space frequency filtering and morphologic space filtering, which are analyzed with their different techniques in regions of interest using the DDSM database.Displaying microcalcifications, which is achieved with the Gaussian high-pass filter in the space frequency and diamond filter, in morphologic space, supported by spectral images (spectrograms) as well as the efficiency, is measured over massive preprocessing images.However, when an improved visualization is presented in processing times, it is observed that the variance in time analyzes a large number of images.Also, although the frequency is faster than morphological space, the morphological filters (Ball filtering) shows a significant result than frequency filters.
Cited by 1Open access
Bogdan Danov, Daniela Gotseva · 2025
In the era of big data, effectively extracting value from huge and complex information flows is key. A set of techniques and paradigms necessary for the successful management and analysis of Big Data are examined. Data preprocessing is examined, and dimensionality reduction techniques, such as PCA and random projections, that optimize data for analysis are also discussed. We look at different storage and access models, as well as data processing paradigms, including batch processing and stream processing.
Cited by 0
Theodoros Iliou, Christos‐Nikolaos Anagnostopoulos, Marina Nerantzaki et al. · 2015
Data preprocessing describes any type of processing methods performed on raw data to prepare it for another processing procedure. Commonly used as a preliminary data mining practice, data preprocessing methods transforms the data into a format that will be more easily and effectively processed for the classification algorithms. In this paper, a novel data preprocessing method is proposed and evaluated in three difficult classification data sets of the well known UCI Repository, in which various classifiers have average performance lower than 75%. The three UCI repository datasets that have been used are the Mammographic masses, Indian Liver and Contraceptive Method. The performance of our proposed data preprocessing method and Principal Component Analysis preprocessing method was evaluated using the 10-fold cross validation method assessing five classification algorithms, Nearest-neighbour classifier (IB1), C4.5 algorithm implementation (J48), Random Forest, Multilayer Perceptron and Rotation Forest, respectively. The classification results are presented and compared analytically. The results indicate that the generated features after our proposed preprocessing method implementation to the original dataset markedly improve the performance of the classification algorithms.
Cited by 48
Wei-He Diao, Xia Mao, Vasile Gui · Electromagnetic waves · 2011
Image preprocessing is commonly used in infrared (IR) small target detection to suppress background clutter and enhance target signature. To evaluate the performance of preprocessing algorithms, two performance metrics, namely PFTN (potential false targets number) decline ratio and BRI (background relative intensity) decline ratio are developed in this paper. The proposed metrics evaluate the performance of given preprocessing algorithm by comparing the qualities of input and output images. The new performance metrics are based on the theories of PFTN and BRI, which describe the quality of IR small target image, by representing the difficulty degree of target detection. Theoretical analysis and experimental results show that the proposed performance metrics can accurately reflect the effect of the image preprocessing stage on reducing false alarms and target shielding. Compared to the traditional metrics, such as signal-to-noise ratio gain and background suppression factor, the new ones are more intuitive and valid.
Cited by 10Open access
Hichem Frigui, Raghu Krishnapuram, D. DeKruger et al. · 2002
We present several fuzzy and robust preprocessing algorithms. The algorithms are specifically designed for target detection in ladar range images. We discuss a fuzzy logic based filtering system that eliminates impulse noise and smoothes images without destroying the details. We present a robust contrast enhancement filter that highlights the target pixels. We also present a background subtraction method based on a robust line fitting algorithm.
Cited by 3
Fang Hsu, Ruei‐Chuan Chang, Richard Chia-Tung Lee · Parallel algorithms and applications · 1993
For an n-gon P, we say P is weakly visible from segment s if any point on P is visible from at least one point of the segment. In this paper, we present an optimal preprocessing algorithm which runs in O(logn) time using O(n) processors under the concurrent read exclusive write parallel random access machine model such that after preprocessing, it takes O(logn) time to test if P is weakly visible from a given segment using a single processor.
Cited by 0