Self-learning Procedure Research Articles

Iterative Methods with Self-Learning for Solving Nonlinear Equations

This paper is devoted to the problem of solving a system of nonlinear equations with an arbitrary but continuous vector function on the left-hand side. By assumption, the values of its components are the only a priori information available about this function. An approximate solution of the system is determined using some iterative method with parameters, and the qualitative properties of the method are assessed in terms of a quadratic residual functional. We propose a self-learning (reinforcement) procedure based on auxiliary Monte Carlo (MC) experiments, an exponential utility function, and a payoff function that implements Bellman’s optimality principle. A theorem on the strict monotonic decrease of the residual functional is proven.

Semi-supervised learning with constrained virtual support vector machines for classification of remote sensing image data

We introduce two semi-supervised models for the classification of remote sensing image data. The models are built upon the framework of Virtual Support Vector Machines (VSVM). Generally, VSVM follow a two-step learning procedure: A Support Vector Machines (SVM) model is learned to determine and extract labeled samples that constitute the decision boundary with the maximum margin between thematic classes, i.e., the Support Vectors (SVs). The SVs govern the creation of so-called virtual samples. This is done by modifying, i.e., perturbing, the image features to which a decision boundary needs to be invariant. Subsequently, the classification model is learned for a second time by using the newly created virtual samples in addition to the SVs to eventually find a new optimal decision boundary. Here, we extend this concept by (i) integrating a constrained set of semi-labeled samples when establishing the final model. Thereby, the model constrainment, i.e., the selection mechanism for including solely informative semi-labeled samples, is built upon a self-learning procedure composed of two active learning heuristics. Additionally, (ii) we consecutively deploy semi-labeled samples for the creation of semi-labeled virtual samples by modifying the image features of semi-labeled samples that have become semi-labeled SVs after an initial model run. We present experimental results from classifying two multispectral data sets with a sub-meter geometric resolution. The proposed semi-supervised VSVM models exhibit the most favorable performance compared to related SVM and VSVM-based approaches, as well as (semi-)supervised CNNs, in situations with a very limited amount of available prior knowledge, i.e., labeled samples.

Deep learning in searching the spectroscopic redshift of quasars

ABSTRACT Studying the cosmological sources at their cosmological rest frames is crucial to track the cosmic history and properties of compact objects. In view of the increasing data volume of existing and upcoming telescopes/detectors, we here construct a 1D convolutional neural network (CNN) with a residual neural network (ResNet) structure to estimate the redshift of quasars in the Sloan Digital Sky Survey IV (SDSS-IV) catalogue from the Data Release 16 Quasar-only (DR16Q) of the extended Baryon Oscillation Spectroscopic Survey on a broad range of signal-to-noise ratios, named FNet. Owing to its 24 convolutional layers and the ResNet structure with different kernel sizes of 500, 200, and 15, FNet is able to discover the local and global patterns in the whole sample of spectra by a self-learning procedure. It reaches the accuracy of 97.0 ${{\ \rm per\ cent}}$ for the velocity difference for redshift, $|\Delta \nu |\lt 6000\, \rm km\, s^{-1}$, and 98.0 ${{\ \rm per\ cent}}$ for $|\Delta \nu |\lt 12\,000\, \rm km\, s^{-1}$, while QuasarNET, which is a standard CNN adopted in the SDSS routine and is constructed of four convolutional layers (no ResNet structure), with kernel sizes of 10, to measure the redshift via identifying seven emission lines (local patterns), fails in estimating redshift of $\sim 1.3{{\ \rm per\ cent}}$ of visually inspected quasars in the DR16Q catalogue, and it gives 97.8 ${{\ \rm per\ cent}}$ for $|\Delta \nu |\lt 6000\, \rm km\, s^{-1}$ and 97.9 ${{\ \rm per\ cent}}$ for $|\Delta \nu |\lt 12\,000\, \rm km\, s^{-1}$. Hence, FNet provides similar accuracy to QuasarNET, but it is applicable for a wider range of SDSS spectra, especially for those missing the clear emission lines exploited by QuasarNET. These properties of FNet, together with the fast predictive power of machine learning, allow FNet to be a more accurate alternative for the pipeline redshift estimator and can make it practical in the upcoming catalogues to reduce the number of spectra to visually inspect.

Twin Self-Learning and Holding Intelligent Control of Fabrication

The paper discusses the problem of managing a three-tier holding with fabrication facilities in subsidiaries. Intellectual management in such a holding is based on the concept of Industry 4.0. At the same time, it is necessary to take into account the undesirable activities of managers playing different roles in the subsidiary. This activity is related to the presence of their own goals, which do not necessarily coincide with the goal of the parent company. With insufficient awareness of this company, these managers can manipulate indicators at subordinate facilities in their favor. In particular, the manager of the subsidiary usually knows better the capabilities of his own fabrication than the higher-level managers. Consequently, such a manager can manipulate the indicators of this fabrication in order to get more positive impetuses (incentives and resources). Such unwanted activity can lead to a decrease in the overall fabrication of the holding. To avoid this, intelligent control is proposed, which includes twin digital self-learning procedures and the formation of positive impetuses. Sufficient conditions have been found for the synthesis of such an intelligent control in which the stochastic possibilities of increasing the holding's fabrication are fully utilized. An example of such intelligent management in the Russian Railways holding is considered.

Abstract 60: Cancer Navigator: A Video-Based Support and Education Tool for Cancer Patients and Their Caregivers

Abstract Purpose: The aim of this study was to identify the informational needs of cancer patients and caregivers among a diverse population in Colombia to inform the development of Cancer Navigator, a video-based tool to help cancer patients and their families understand and navigate their treatment. Methods: Seven focus group discussions (FGDs) with n=47 participants and n=80 in-depth interviews (IDIs) were conducted to understand the needs of cancer patients and caregivers regarding their cancer diagnosis, treatment and rehabilitation in the Colombian healthcare system. The information was analyzed using thematic analysis, and informed the development of a series of 18 short infographic and personal narrative videos covering different cancer topics, delivered by local clinicians and survivors; along with a resource guide of local services for food, housing and transportation. A pilot study was conducted to evaluate pre-post depression and anxiety among 22 caregivers and 18 cancer patients at Hospital Méderi (Bogotá, Colombia). Results: Themes/topics identified from the interviews included: treatment effects; cancer etiology; nutrition, physical activity recommendations; psychological/spiritual/emotional support; alternative therapies; warning signs; self-learning cancer tools and administrative procedures. The pilot study showed a mild decrease in anxiety and distress immediately after viewing three videos. The proportion of patients with anxiety levels above normal (>7 on Hospital Anxiety and Depression Scale, HADS) changed from 27% to 15% after viewing (p=0.17), and distress (score >5 on the distress thermometer) decreased from 35% to 22% (p=0.22). Conclusion: The Cancer Navigator video-based educational/support tool shows promise for reducing anxiety and distress among cancer patients and caregivers in Colombia. Next steps will be to test the feasibility and acceptability of the videos among patients, caregivers and healthcare providers at three additional oncology hospitals in Bogotá in 2021. Citation Format: Ana Pedraza-Flechas, Carolyn Taylor, Erika Barrera-Suárez, Luis Perez, Jesica Soto, Alejandra Arias, Andres Patiño, Vera Perales, Ángela Pinzón-Rondon, Ángela Ruíz-Sternberg, Anna Cabanes, Phuongthao Le. Cancer Navigator: A Video-Based Support and Education Tool for Cancer Patients and Their Caregivers [abstract]. In: Proceedings of the 9th Annual Symposium on Global Cancer Research; Global Cancer Research and Control: Looking Back and Charting a Path Forward; 2021 Mar 10-11. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2021;30(7 Suppl):Abstract nr 60.

Industrial research on evolution and prediction of hardwood color.

Color prediction in dyed wood is a difficult task since it involves the analysis of light propagation through a complex media where scattering and absorption processes are present. Kubelka-Munk-based models are usually proposed to make those predictions. Here, an oak wood color prediction tool is presented with the Kubelka-Munk theory and self-learning procedures as the basis of the model. Color prediction lies on the joint contribution of both the dying material and the wood substrate, each characterized by their previously obtained colorimetric and spectral properties. An identification of wood and dyes through the study of their optical properties is shown, from which the necessary parameters are obtained for the different applications. The model allows us to predict with good accuracy the resulting color in wood through the L∗C∗h∘ coordinates when mixing either water or solvent-based dyes in different proportions for dying a wood substrate. Furthermore, the influences of applying dye mixtures either by hand with a brush or by machine with a roller coating and also that of varnishing are studied.

SELF-LEARNING ONTOLOGY FOR INSTANCE SEGMENTATION OF 3D INDOOR POINT CLOUD

Abstract. Automation in point cloud data processing is central for efficient knowledge discovery. In this paper, we propose an instance segmentation framework for indoor buildings datasets. The process is built on an unsupervised segmentation followed by an ontology-based classification reinforced by self-learning. We use both shape-based features that only leverages the raw X, Y, Z attributes as well as relationship and topology between voxel entities to obtain a 3D structural connectivity feature describing the point cloud. These are then used through a planar-based unsupervised segmentation to create relevant clusters constituting the input of the ontology of classification. Guided by semantic descriptions, the object characteristics are modelled in an ontology through OWL2 and SPARQL to permit structural elements classification in an interoperable fashion. The process benefits from a self-learning procedure that improves the object description iteratively in a fully autonomous fashion. Finally, we benchmark the approach against several deep-learning methods on the S3DIS dataset. We highlight full automation, good performances, easy-integration and a precision of 99.99% for planar-dominant classes outperforming state-of-the-art deep learning.

Self-Learning Method for Construction of Analytical Interatomic Potentials to Describe Laser-Excited Materials.

Large-scale simulations using interatomic potentials provide deep insight into the processes occurring in solids subject to external perturbations. The atomistic description of laser-induced ultrafast nonthermal phenomena, however, constitutes a particularly difficult case and has so far not been possible on experimentally accessible length scales and timescales because of two main reasons: (i)abinitio simulations are restricted to a very small number of atoms and ultrashort times and (ii)simulations relying on electronic temperature- (T_{e}) dependent interatomic potentials do not reach the necessary abinitio accuracy. Here we develop a self-learning method for constructing T_{e}-dependent interatomic potentials which permit ultralarge-scale atomistic simulations of systems suddenly brought to extreme nonthermal states with density-functional theory (DFT) accuracy. The method always finds the global minimum in the parameter space. We derive a highly accurate analytical T_{e}-dependent interatomic potential Φ(T_{e}) for silicon that yields a remarkably good description of laser-excited and -unexcited Si bulk and Si films. Using Φ(T_{e}) we simulate the laser excitation of Si nanoparticles and find strong damping of their breathing modes due to nonthermal melting.

Yin-Yang firefly algorithm based on dimensionally Cauchy mutation

Firefly algorithm (FA) is a classical and efficient swarm intelligence optimization method and has a natural capability to address multimodal optimization. However, it suffers from premature convergence and low stability in the solution quality. In this paper, a Yin-Yang firefly algorithm (YYFA) based on dimensionally Cauchy mutation is proposed for performance improvement of FA. An initial position of fireflies is specified by the good nodes set (GNS) strategy to ensure the spatial representativeness of the firefly population. A designed random attraction model is then used in the proposed work to reduce the time complexity of the algorithm. Besides, a key self-learning procedure on the brightest firefly is undertaken to strike a balance between exploration and exploitation. The performance of the proposed algorithm is verified by a set of CEC 2013 benchmark functions used for the single objective real parameter algorithm competition. Experimental results are compared with those of other the state-of-the-art variants of FA. Nonparametric statistical tests on the results demonstrate that YYFA provides highly competitive performance in terms of the tested algorithms. In addition, the application in constrained engineering optimization problems shows the practicability of YYFA algorithm.

Self-Learning and Efficient Health-Status Analysis for a Core Router System

The health status of core router systems needs to be analyzed efficiently in order to ensure high reliability and timely error recovery. Although a large amount operational data is collected from core routers, due to high computational complexity and expensive labor cost, only a small part of this data is labeled by experts. The lack of labels is an impediment toward the adoption of supervised learning. We present an iterative self-learning procedure for assessing the health status of a core router. This procedure first computes a representative feature matrix to capture different characteristics of time-series data. Not only statistical-modeling-based features are computed from three general categories but also a recurrent neural network-based autoencoder is utilized to capture a wider range of hidden patterns. Moreover, both minimum-redundancy–maximum-relevance (mRMR) method and fully connected feedforward autoencoder are applied to further reduce dimensionality of extracted feature matrix. Hierarchical clustering is then utilized to infer labels for the unlabeled dataset. Finally, a classifier is built and iteratively updated using both labeled and unlabeled dataset. Field data collected from a set of commercial core routers are used to experimentally validate the proposed health-status analyzer. The experimental results show that the proposed feature-based self-learning health analyzer achieves higher precision and recall than the traditional supervised health analyzer as well the currently deployed rule-based health analyzer. Moreover, it achieves better performance than the three anomaly detection baseline methods under the transformed binary classification scenario.

Virtual Support Vector Machines with self-learning strategy for classification of multispectral remote sensing imagery

We follow the idea of learning invariant decision functions for remote sensing image classification with Support Vector Machines (SVM). To do so, we generate artificially transformed samples (i.e., virtual samples) from available prior knowledge. Labeled samples closest to the separating hyperplane with maximum margin (i.e., the Support Vectors) are identified by learning an initial SVM model. The Support Vectors are used for generating virtual samples by perturbing the features to which the model should be invariant. Subsequently, the model is relearned using the Support Vectors and the virtual samples to eventually alter the hyperplane with maximum margin and enhance generalization capabilities of decision functions. In contrast to existing approaches, we establish a self-learning procedure to ultimately prune non-informative virtual samples from a possibly arbitrary invariance generation process to allow for robust and sparse model solutions. The self-learning strategy jointly considers a similarity and margin sampling constraint. In addition, we innovatively explore the invariance generation process in the context of an object-based image analysis framework. Image elements (i.e., pixels) are aggregated to image objects (as represented by segments/superpixels) with a segmentation algorithm. From an initial singular segmentation level, invariances are encoded by varying hyperparameters of the segmentation algorithm in terms of scale and shape. Experimental results are obtained from two very high spatial resolution multispectral data sets acquired over the city of Cologne, Germany, and the Hagadera Refugee Camp, Kenya. Comparative model accuracy evaluations underline the favorable performance properties of the proposed methods especially in settings with very few labeled samples.

Designing adaptive information models for production management

A methodology of open design of adaptive information models for production management under uncertainty, based on a cognitive approach and taking into account the human factor, is developed. This methodology is based on the idea of designing a variety of adaptive information models based on a single basic model. In this case, a variety of adaptation procedures that are publicly available can be used. As an example, information models based on self-learning procedures are considered. The results of the operation of these models are adaptive parameters (norms) and production estimates. When designing such models, it can be used many of the formal self-learning procedures described in the scientific and technical literature. As an example, the self-learning information model for production management based on a stochastic approximation designed. Sufficient conditions for its correctness are found, which ensure the unfolding of the production potential taking into account the human factor. The application of the self-learning information model is illustrated by the example of wagon-repair production of large-scale corporation Russian Railways.

Robust real-time heart rate prediction for multiple subjects from facial video using compressive tracking and support vector machine.

Remote monitoring of vital physiological signs allows for unobtrusive, nonrestrictive, and noncontact assessment of an individual's health. We demonstrate a simple but robust image photoplethysmography-based heart rate (HR) estimation method for multiple subjects. In contrast to previous studies, a self-learning procedure of tech was developed in our study. We improved compress tracking algorithm to track the regions of interest from video sequences and used support vector machine to filter out potentially false beats caused by variations in the reflected light from the face. The experiment results on 40 subjects show that the absolute value of mean error reduces from 3.6 to . We further explore experiments for 10 subjects simultaneously, regardless of the videos at a resolution of 600 by 800, the HR is predicted real-time and the results reveal modest but significant effects on HR prediction.

Energy assessment method towards low-carbon energy schools

Within the building sector, schools have a major social responsibility because of their educational purpose. With the aim of providing a novel energy modelling process to evaluate the real energy performance of school buildings and potential energy savings, minimizing input data collection, this paper presents a new energy assessment method to support the decision-making process towards low-carbon energy schools. The novelty of this method is based on the integration of a coherent set of assumptions and procedures with respect to boundary conditions of schools, derived from their modular basis, common building configuration and space uses, and the model's iterative calibration procedure, based on real building performance, which achieves a high final accuracy. With a reduced set of inputs, hourly dynamic simulations can be performed. The method, integrated in the SchoolEnergy-ACT tool, was tested in two pilot schools. The results show that the method properly fits to schools' energy consumption profiles with high accuracy levels through calibration processes using energy bills, ensuring its feasibility for simulating the energy performance of schools, enabling comparison with other buildings, and informing end-users of potential energy savings. The accuracy of the model can be improved with an iterative, self-learning procedure and detailed energy data.

A neuro-fuzzy Kohonen network for data stream possibilistic clustering and its online self-learning procedure

A task of Data Stream Fuzzy Clustering is considered when data is processed sequentially under a priori uncertainty conditions about both a number of clusters and a degree of clusters’ overlapping. A modified two-layer neuro-fuzzy Kohonen network is used for solving the possibilistic fuzzy clustering tasks. This system tunes centers’ coordinates and membership levels of every pattern to clusters during the self-learning procedure and automatically increases a number of neurons during data processing. A distinguishing feature of the proposed approach is its computational simplicity due to the fact that a recurrent modification of the possibilistic fuzzy clustering procedure is used for tuning the network’s parameters. The proposed neuro-fuzzy system is based on the concepts of evolving systems of Computational Intelligence, the recurrent optimization, the possibilistic fuzzy clustering, and Data Stream Mining.

Predicting Waterflooding Performance in Low-Permeability Reservoirs With Linear Dynamical Systems

Summary Several interwell connectivity models such as multiple linear regression (MLR) and the capacitance model (CM) have been proposed to model waterflooding performance in high-permeability reservoirs on the basis of observed production data. However, the existing methods are not effective at characterizing the behavior of transient flows that are prevalent in low-permeability reservoirs. This paper presents a novel dynamic waterflooding model that is based on linear dynamical systems (LDSs) to characterize the injection/production relationships in an oil field during both stationary and nonstationary production phases. We leverage a state-space model (SSM), in which the changing rates of control volumes between injector/producer pairs in the reservoir of interest serve as time-varying hidden states, depending on the reservoir condition. Thus, the model can better characterize the transient dynamics in low-permeability reservoirs. We propose a self-learning procedure for the model to train its parameters as well as the evolution of the hidden states only on the basis of past observations of injection and production rates. We tested the LDS method in comparison with the state-of-the-art CM method in a wide range of synthetic reservoir models including both high-permeability and low-permeability reservoirs, as well as various dynamic scenarios involving varying bottomhole pressure (BHP) of producers, injector shut-ins, and reservoirs of larger scales. We also tested LDS on the real production data collected from Changqing oil field containing low-permeability formations. Testing results demonstrate that an LDS significantly outperforms CM in terms of modeling and predicting waterflooding performance in low-permeability reservoirs and various dynamic scenarios.

A New Data Analysis System to Quantify Associations between Biochemical Parameters of Chronic Kidney Disease-Mineral Bone Disease.

BackgroundIn hemodialysis patients, deviations from KDIGO recommended values of individual parameters, phosphate, calcium or parathyroid hormone (PTH), are associated with increased mortality. However, it is widely accepted that these parameters are not regulated independently of each other and that therapy aimed to correct one parameter often modifies the others. The aim of the present study is to quantify the degree of association between parameters of chronic kidney disease and mineral bone disease (CKD-MBD).MethodsData was extracted from a cohort of 1758 adult HD patients between January 2000 and June 2013 obtaining a total of 46.141 records (10 year follow-up). We used an advanced data analysis system called Random Forest (RF) which is based on self-learning procedure with similar axioms to those utilized for the development of artificial intelligence. This new approach is particularly useful when the variables analyzed are closely dependent to each other.ResultsThe analysis revealed a strong association between PTH and phosphate that was superior to that of PTH and Calcium. The classical linear regression analysis between PTH and phosphate shows a correlation coefficient is 0.27, p<0.001, the possibility to predict PTH changes from phosphate modification is marginal. Alternatively, RF assumes that changes in phosphate will cause modifications in other associated variables (calcium and others) that may also affect PTH values. Using RF the correlation coefficient between changes in serum PTH and phosphate is 0.77, p<0.001; thus, the power of prediction is markedly increased. The effect of therapy on biochemical variables was also analyzed using this RF.ConclusionOur results suggest that the analysis of the complex interactions between mineral metabolism parameters in CKD-MBD may demand a more advanced data analysis system such as RF.

Context-aware discovery of visual co-occurrence patterns.

Once an image is decomposed into a number of visual primitives, e.g., local interest points or regions, it is of great interests to discover meaningful visual patterns from them. Conventional clustering of visual primitives, however, usually ignores the spatial and feature structure among them, thus cannot discover high-level visual patterns of complex structure. To overcome this problem, we propose to consider spatial and feature contexts among visual primitives for pattern discovery. By discovering spatial co-occurrence patterns among visual primitives and feature co-occurrence patterns among different types of features, our method can better address the ambiguities of clustering visual primitives. We formulate the pattern discovery problem as a regularized k-means clustering where spatial and feature contexts are served as constraints to improve the pattern discovery results. A novel self-learning procedure is proposed to utilize the discovered spatial or feature patterns to gradually refine the clustering result. Our self-learning procedure is guaranteed to converge and experiments on real images validate the effectiveness of our method.

A diagnostic logic for preventing structural failure in concrete displacing booms

Abstract During recent decades safety in the workplace has become more and more important. For this reason, systems for the diagnosis of possible failures represent a particularly attractive solution. In this paper a health monitoring algorithm for a specific kind of construction machinery (concrete displacing booms) is proposed. It is based on the knowledge of the geometrical and dynamic parameters of the boom, estimated through a stand-alone self-learning procedure. The proposed algorithm is able to signal alarms due to the most common failure conditions, such as overload or crack propagation. In detail, the crack propagation is detected through an Extended Kalman Filter (which identifies the variation of the boom natural frequencies during its lifetime), while the overload is estimated through force sensors mounted on the actuators. The results referred to this algorithm, numerically simulated and experimentally reproduced on a specific test rig, are presented and discussed.

License Plate Localization and Character Segmentation With Feedback Self-Learning and Hybrid Binarization Techniques

License plate localization (LPL) and character segmentation (CS) play key roles in the license plate (LP) recognition system. In this paper, we dedicate ourselves to these two issues. In LPL, histogram equalization is employed to solve the low-contrast and dynamic-range problems; the texture properties, e.g., aspect ratio, and color similarity are used to locate the LP; and the Hough transform is adopted to correct the rotation problem. In CS, the hybrid binarization technique is proposed to effectively segment the characters in the dirt LP. The feedback self-learning procedure is also employed to adjust the parameters in the system. As documented in the experiments, good localization and segmentation results are achieved with the proposed algorithms.