Search

Earthquakes are hazardous natural disasters, and they may cause severe damage and losses where they occur. In addition to their devastating effects, they may trigger following disasters like tsunamis and fires. Post-earthquake fires are known as the most dangerous secondary disasters and generally cause much more damage than the damage caused by the earthquake itself. The difficulty in determining and responding to ignition sources, the lack of equipment and workforce, and obstacles like collapsed buildings that block the ways to reach fires may cause catastrophic disasters after an earthquake. In recent years, Unmanned Aerial Vehicle technologies (UAVs) have shown promising performance in post-disaster response operations. Parallel to technological improvements, they have been used for many purposes, like fire-fighting, victim location detection, base station support, and material distribution in disaster areas. To manage a possible response and improve the performance of UAVs in post-earthquake fire areas, it is crucial to be prepared in advance. This study proposes an artificial neural network-based clustering approach for unmanned aerial vehicle use in post-earthquake fire areas. After conducting a detailed literature review covering post-earthquake fires, usage of UAVs in disasters, and some aspects of Self Organizing Maps, the methodology used for clustering the neighborhoods regarding their post-earthquake fire risk similarities is introduced. A real-life application is carried out to identify and cluster the regions and provide preliminary information to the decision-makers on possible interventions. Neighborhoods of Tuzla district, one of the riskiest districts in terms of post-earthquake fires in Istanbul, are clustered with Self-Organizing Maps (SOM). In a possible post-earthquake fire disaster, the Tuzla district can be divided into three areas, and UAVs can be organized more efficiently and quickly based on this cluster information. The results of this real-life application can guide decision-makers by showing which regions have similarities for UAV response in possible post-earthquake fires and where they can be intervened together. The authorities can benefit from the findings of this study while preparing disaster plans, intervention actions, and post-disaster humanitarian activities. [ABSTRACT FROM AUTHOR]

Due to the everchanging and evergrowing nature of programming technologies, the gap between the programming industry's needs and the educational capabilities of both formal and informal educational environments has never been wider. However, the need to learn computer programming has never been greater, regardless of the motivation behind it. The number of programming concepts to be taught is increasing over time, while the amount of time available for education and training usually remains the same. The objective of this research was to analyze the source codes used in many educational systems to teach fundamental programming concepts to learners, regardless of their prior experience in programming. A total of 25 repositories containing 3882 Python modules were collected for the analysis. Through self-organization of the collected content, we obtained very compelling results about code structure, distribution, and differences. Based on those results, we concluded that Self-Organizing Maps are a powerful tool for both content and knowledge management, because they can highlight problems in the curriculum's density as well as transparently indicate which programming concepts it has successfully observed and learned to recognize. Based on the level of transparency exhibited by Self-Organizing Maps, it is safe to say that they could be used in future research to enhance both human and machine learning of computer programming. By achieving this level of transparency, such an Artificial Intelligence system would be able to assist in overall computer programming education by communicating what should be taught, what needs to be learned, and what is known. [ABSTRACT FROM AUTHOR]

Tailored analytics play a key role in the successful delivery of cultural content to huge and diverse groups. Primarily the latter depends on a number of information retrieval factors determining user experience quality, most prominently precision, recall, and timing. These imply that cultural analytics should be designed with strong predictive power. In turn, the latter relies heavily on the clustering of the system user base. A self organizing map is a neural network architecture trained in an unsupervised way through a modified Hebbian rule to couple distances between two distinct spaces such that a manifold in the high dimensional space is projected smoothly to the lower dimensional one. The twofold focus of this work is the development of a tensor user distance metric for SOMs as well as the inclusion of behavioral attributes therein, both aiming at additional descriptive power and clustering flexibility. As a concrete example, the proposed SOMs are applied to data taken from a cultural content delivery system. The proposed methodology is evaluated based on a scoring method assessing both complexity and clustering quality criteria, including the number of epochs, the average cluster distance, and the topological error, with encouraging results. [ABSTRACT FROM AUTHOR]

Self Organizing Maps (SOM) are unsupervised neural networks suited for visualisation purposes and clustering analysis. This study uses SOM to solve a software engineering problem: detecting the most important (key) classes from software projects. Key classes are meant to link the most valuable concepts of a software system and in general these are found in the solution documentation. UML models created in the design phase become deprecated in time and tend to be a source of confusion for large legacy software. Therefore, developers try to reconstruct class diagrams from the source code using reverse engineering. However, the resulting diagram is often very cluttered and difficult to understand. There is an interest for automatic tools for building concise class diagrams, but the machine learning possibilities are not fully explored at the moment. This paper proposes two possible algorithms to transform SOM in a classification algorithm to solve this task, which involves separating the important classes - that should be on the diagrams - from the others, less important ones. Moreover, SOM is a reliable visualization tool which able to provide an insight about the structure of the analysed projects. [ABSTRACT FROM AUTHOR]

The boundary layer (BL) profile over the coastal plain of Israel, Eastern Mediterranean (EM), varies considerably during winter. Although in the context of air pollution, the characteristics of the BL height (BLH) was intensively investigated, a quantitative classification of the BL profile regimes has not been performed. Here, we seek to reveal the dominant, recurring regimes of the BL profiles, their quantitative characteristics and links to regional synoptic‐scale patterns. An objective unsupervised classification of winter BL radiosonde profiles is performed for the first time by multi‐parameter self‐organizing map (SOM) analysis. The analysis uses high‐resolution, 12:00‐UTC data of wind, temperature, humidity and pressure measurements during December to February 2007–2018, and yields 30 distinct profile regimes. Composite analysis using ERA5 reanalysis suggests strong association between the profile regimes and synoptic weather systems and highlights four groups: (a) Deep winter cyclones with strong westerly wind and precipitation; (b) Strong surface anticyclones and Red Sea troughs (RST) with a mid‐tropospheric ridge, moderate dry easterly wind and extreme temperatures. (c) Moderate pressure gradients under shallow cyclones, anticyclone to the west and RST to the east of Israel. (d) Active RSTs, accompanied by upper‐tropospheric trough/cutoff low and heavy precipitation. For the first time, general objective classification observes the active RST without requiring specific criteria. Consistent with previous knowledge, the new classification exhibits distinct categories of thermal stability, BLH and turbulence. Importantly, we show that the automatic objective classification of profile data from a single station can be a sensitive discriminator of winter synoptic regimes in the EM, and therefore explains the variability of the BL profile. It facilitates the study of the interaction between the BL and the free troposphere and may improve the prediction of air pollution or future BL profile regimes based on long time series from historical data or climate models. [ABSTRACT FROM AUTHOR]

Extractive text summarization yields the sensitive parts of the document by neglecting the irrelevant and redundant information. In this paper, we propose a new strategy for extractive single-document summarization in Malayalam. Initially, entity recognition is done, followed by relevance analysis is made based on some context-aware features. The scored sentences are then clustered using self-organizing maps (SOM) and from these clusters, relevant sentences are extracted out based on the proposed algorithm. Both theoretical and practical evaluations are done to analyze the implemented system. In theoretical evaluation, gradient calculations of relevance equations are used to know that which of these sentence scoring features are contributing more. The relevance equation is optimized with the help of Lagrange's multiplier. The complexity analysis of the proposed algorithms is also performed. In practical evaluation, the system compared with online and offline summarizers upon metrics like precision, recall, and F-measure. The system is tested through a non-clustering approach also in order to analyze the impact of clustering used in our work. Some existing strategies like question game evaluation, sentence rank evaluation, and keyword association are also done to evaluate the different parameters like the relevance of sentences, important entity words, etc. [ABSTRACT FROM AUTHOR]

The paper presents the application of the IMGAMO (immune game theory multiobective algorithm) in the optimal design of electrothermal microactuators. Several numerical tests on the mathematical benchmark test functions were performed, showing the superiority of the IMGAMO, when a large number of criteria are considered, in comparison to other multiobjective optimizers. A parametric numerical model of an electrothermal microactuaror was developed and verified. Six functionals, which depend on various thermal and mechanical quantities of the microactuator, were proposed, formulated and numerically implemented. These functionals represent real requirements asked of microactuators. The boundary-value problem of an electro-thermo-mechanical field was solved multiple times during the course of optimization by way of the finite element method (FEM). A numerical example of multiobjective optimization of chevron-type electrothermal actuators is included in the paper. Representation of the multi-dimensional Pareto fronts by means of scatter plot matrices, aided by self-organizing maps (SOMs), is presented. The novel method of selecting interesting, compromise-solutions is proposed and described. [ABSTRACT FROM AUTHOR]

In emergency scenarios service vehicles must identify potential route(s) and use the best available route. However, route identification requires intelligent decision-support systems which generally use non-traditional approaches with tools characterised by flexible non-hierarchical structures. Conventional models using group decision-support systems have been applied; however, when used in smart urban environments, emergency response services have limitations in their ability to identify unobstructed paths (routes) in dynamic operating environments. In this paper we introduce a novel path planning method for autonomous vehicle control in emergency situations. The proposed model uses self-organizing maps in an integrated Spiral STC algorithm termed the: Hybrid SOM-Spiral STC model which uses hedge algebras and Kansei evaluation in group decision-support. The proposed model has been designed to quantify qualitative factors using sensor derived data processed with human sensibilities and preferences in emergency decision support. The experimental results show that the proposed model achieves significant improvements in group decision-support under dynamic uncertainty. We posit that our novel approach holds the prospect of improvements in the provision of emergency services. [ABSTRACT FROM AUTHOR]

Highlights • A self-adjustable domain adaptation (SADA) strategy is devised against overfitting. • SADA builds a vital sign dataset of ECG and IR-UWB radar data by actual records. • SADA devises the SOM-based ECG clustering by transfer learning to fuse multi data. • SADA broadens the application of domain adaptation algorithms by importing OC-SVM. Abstract To enhance electrocardiogram (ECG) monitoring systems in personalized detections, deep neural networks (DNNs) are applied to overcome individual differences by periodical retraining. As introduced previously [4] , DNNs relieve individual differences by fusing ECG with impulse radio ultra-wide band (IR-UWB) radar. However, such DNN-based ECG monitoring system tends to overfit into personal small datasets and is difficult to generalize to newly collected unlabeled data. This paper proposes a self-adjustable domain adaptation (SADA) strategy to prevent from overfitting and exploit unlabeled data. Firstly, this paper enlarges the database of ECG and radar data with actual records acquired from 28 testers and expanded by the data augmentation. Secondly, to utilize unlabeled data, SADA combines self organizing maps with the transfer learning in predicting labels. Thirdly, SADA integrates the one-class classification with domain adaptation algorithms to reduce overfitting. Based on our enlarged database and standard databases, a large dataset of 73200 records and a small one of 1849 records are built up to verify our proposal. Results show SADA's effectiveness in predicting labels and increments in the sensitivity of DNNs by 14.4% compared with existing domain adaptation algorithms. [ABSTRACT FROM AUTHOR]

In this paper a particular emphasis is given on developing a character recognition system using SciLab, a free and open source computing software and is most promising alternative to MatLab. In the proposed work a character recognition system to extract printed text from an image is developed using Kohenen self organizing maps (SOM) based retrieval system. SOM being an unsupervised method of training has a superior feature extracting property. Samples of same characters which are oriented at same angle but with different size, color and fonts are used. After calculation of certain topological and geometrical properties of a character it is classified and recognized. With self organizing map together with K means clustering algorithm using SciLab software, the system has achieved a remarkable accuracy of 99% to 100%, when tested for various text input images. [ABSTRACT FROM AUTHOR]

High dimensional data visualization is one of the main tasks in the field of data mining and pattern recognition. The self organizing maps (SOM) is one of the topology visualizing tool that contains a set of neurons that gradually adapt to input data space by competitive learning and form clusters. The topology preservation of the SOM strongly depends on the learning process. Due to this limitation one cannot guarantee the convergence of the SOM in data sets with clusters of arbitrary shape. In this paper, we introduce Constrained SOM (CSOM), the new version of the SOM by modifying the learning algorithm. The idea is to introduce an adaptive constraint parameter to the learning process to improve the topology preservation and mapping quality of the basic SOM. The computational complexity of the CSOM is less than those with the SOM. The proposed algorithm is compared with similar topology preservation algorithms and the numerical results on eight small to large real-world data sets demonstrate the efficiency of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Recent estimations considered that up to 40% of oil spills in the US governmental jurisdiction remained unidentified. Many circumstances may be behind this striking figure, one of them the difficulties in the analytical characterization of the test aliquots and the subsequent extraction of information. In this paper a simple approach to handle complex datasets (organized as oils × weathering aliquots × analytical variables) is proposed. It considers self-organizing maps to gather information about: (i) the samples, and, so, whether the oils can be differentiated; (ii) the weathering of the oils; and (iii) the analytical variables (in this case, mid-IR spectral indexes) that characterize each oil and weathering. It was verified that the oils can be differentiated, even after their weathering. Besides, a quite good ordering of the aliquots as a function of time was revealed. Considering the two different arrangements that were studied ( I × J × K or J × I × K ) the results obtained are totally similar. The major difference was that when the ( I × J × K ) arrangement; i.e., oils × weathering aliquots × IR indexes, was considered, the final PCA (made with the MOLMAP-scores) discriminated immediately between the different products. Considering the ( J × I × K ) ordering; i.e., weathering aliquots × oils × IR indexes, the PCA emphasized the different stages of weathering. [ABSTRACT FROM AUTHOR]

The main goal of this study is a comparison of two different methods of pattern recognition. The first, Principal Component Analysis (PCA), is a method frequently used in climatology. The second, Self Organizing Maps (SOM), is a relatively new and efficient method based on Artificial Neural Networks (ANN). In order to compare the two methodologies, two teleconnection patterns were chosen, the North Atlantic Oscillation (NAO), which mostly affects the climate of Western Europe, and the North Sea–Caspian Pattern (NCP), mainly affecting eastern Mediterranean and the Balkan Peninsula. The teleconnection patterns are studied for winter 500 hPa geopotential height anomalies over the broader Europe area. The secondary objective of the study is to evaluate the ECHAM5/MPI General Circulation Model (GCM) in representing the two teleconnection patterns for a reference (1971–2000) and a future (2071–2100) period. The evaluation of the reference period is done comparing the simulated data to the NCEP/NCAR reanalysis data. The future period is studied to examine whether the current dominant circulation patterns change or not during the last 30 years of the 21st century. According to the results, both PCA and SOM methodologies capture the main variability mode over the study area, represented by the NAO pattern, but SOM is capable of capturing even less pronounced patterns, such as the NCP. In the future simulations, the atmospheric circulation during winter seems to be more pronounced with stronger NAO and NCP teleconnection patterns. [ABSTRACT FROM AUTHOR]

Multivariate adaptive regression splines (MARS) has become a popular data mining (DM) tool due to its flexible model building strategy for high dimensional data. Compared to well-known others, it performs better in many areas such as finance, informatics, technology and science. Many studies have been conducted on improving its performance. For this purpose, an alternative backward stepwise algorithm is proposed through Conic-MARS (CMARS) method which uses a penalized residual sum of squares for MARS as a Tikhonov regularization problem. Additionally, by modifying the forward step of MARS via mapping approach, a time efficient procedure has been introduced by S-FMARS. Inspiring from the advantages of MARS, CMARS and S-FMARS, two hybrid methods are proposed in this study, aiming to produce time efficient DM tools without degrading their performances especially for large datasets. The resulting methods, called SMARS and SCMARS, are tested in terms of several performance criteria such as accuracy, complexity, stability and robustness via simulated and real life datasets. As a DM application, the hybrid methods are also applied to an important field of finance for predicting interest rates offered by a Turkish bank to its customers. The results show that the proposed hybrid methods, being the most time efficient with competing performances, can be considered as powerful choices particularly for large datasets. [ABSTRACT FROM AUTHOR]

Self Organizing Maps (SOMs) are widely used neural networks for classification or visualization of large datasets. Like many neural network simulations, implementations of the SOM algorithm need a scan of all the neural units in order to simulate the work of a parallel machine. This paper reports a new learning algorithm that speeds up the training of a SOM with a little loss of the performance on many quality tests. The very low computation time, means that this algorithm can be used as a fast visualization tool for large multidimensional datasets. [ABSTRACT FROM AUTHOR]

Abstract: Modeling and recognizing spatiotemporal, as opposed to static input, is a challenging task since it incorporates input dynamics as part of the problem. The vast majority of existing methods tackle the problem as an extension of the static counterpart, using dynamics, such as input derivatives, at feature level and adopting artificial intelligence and machine learning techniques originally designed for solving problems that do not specifically address the temporal aspect. The proposed approach deals with temporal and spatial aspects of the spatiotemporal domain in a discriminative as well as coupling manner. Self Organizing Maps (SOM) model the spatial aspect of the problem and Markov models its temporal counterpart. Incorporation of adjacency, both in training and classification, enhances the overall architecture with robustness and adaptability. The proposed scheme is validated both theoretically, through an error propagation study, and experimentally, on the recognition of individual signs, performed by different, native Greek Sign Language users. Results illustrate the architecture’s superiority when compared to Hidden Markov Model techniques and variations both in terms of classification performance and computational cost. [Copyright &y& Elsevier]

Abstract: The extreme wave climate is of paramount importance for: (i) off-shore and coastal engineering design, (ii) ship design and maritime transportation, or (iii) analysis of coastal processes. Identifying the synoptic patterns that produce extreme waves is necessary to understand the wave climate for a specific location. Thus, a characterization of these weather patterns may allow the study of the relationships between the magnitude and occurrence of extreme wave events and the climate system. The aim of this paper is to analyze the interannual variability of extreme wave heights. For this purpose, we present a methodological framework and its application to an area over the North East (NE) Atlantic Ocean. The climatology in the NE Atlantic is analyzed using the self-organizing maps (SOMs). The application of this clustering technique to monthly mean sea level pressure fields provides a continuum of synoptic categorizations compared with discrete realizations produced through most traditional methods. The extreme wave climate has been analyzed by means of monthly maxima of the significant wave height (SWH) in several locations over the NE Atlantic. A statistical approach based on a time-dependent generalized extreme value (GEV) distribution has been applied. The seasonal variation was characterized and, afterwards, the interannual variability was studied throughout regional pressure patterns. The anomalies of the 50-year return level estimates of SWH, due to interannual variability have been projected into the weather types of SOM. It provides a comprehensive visual representation, which relates the weather type with the positive or negative contribution to extreme waves over the selected locations. [Copyright &y& Elsevier]

Abstract: Kohonen maps and Counterpropagation Neural Networks are two of the most popular learning strategies based on Artificial Neural Networks. Kohonen Maps (or Self Organizing Maps) are basically self-organizing systems which are capable to solve the unsupervised rather than the supervised problems, while Counterpropagation Artificial Neural Networks are very similar to Kohonen maps, but an output layer is added to the Kohonen layer in order to handle supervised modelling. Recently, the modifications of Counterpropagation Artificial Neural Networks allowed introducing new supervised neural network strategies, such as Supervised Kohonen Networks and XY-fused Networks. In this paper, the Kohonen and CP-ANN toolbox for MATLAB is described. This is a collection of modules for calculating Kohonen maps and derived methods for supervised classification, such as Counterpropagation Artificial Neural Networks, Supervised Kohonen Networks and XY-fused Networks. The toolbox comprises a graphical user interface (GUI), which allows the calculation in an easy-to-use graphical environment. It aims to be useful for both beginners and advanced users of MATLAB. The use of the toolbox is discussed here with an appropriate practical example. [Copyright &y& Elsevier]

Abstract: The regional dynamics of industrial fisheries within Large Marine Ecosystems (LMEs) boundaries were investigated by means of a historical-descriptive approach. Landings data from the Sea Around Us Project database were used to detect trends in total yields and variations in landings composition by functional groups over time. The temporal and spatial scales covered by this study allowed general issues to be addressed such as the detection of recurrent patterns and synchronies in fisheries landings. An unsupervised artificial neural network, Self Organizing Map (SOM), is used as a tool to analyze fisheries landings composition variation over five decades in 51 LMEs all over the world. From the historical analysis of “fishing behaviors” within LMEs a broad distinction between two main types of fisheries emerged: (1) small and medium pelagics fisheries, with stable compositions or cyclic behaviors, occurred in LMEs which share common productive features, despite different geographical locations and (2) demersal fisheries, which are more affected by economic drivers and tend to concentrate in LMEs in the Northern Hemisphere. Our analysis can be regarded as a first step towards the challenging scope of describing the relative influence of environmental and economic drivers on exploited ecosystems. [Copyright &y& Elsevier]

Abstract: This paper introduces a technique that uses frequent pattern mining and SOM techniques to identify, group and analyse trends in sequences of time stamped social networks so as to identify “interesting” trends. In this study, trends are defined in terms of a series of occurrence counts associated with frequent patterns that may be identified within social networks. Typically a large number of frequent patterns, and by extension a large number of trends, are discovered. Thus, to assist with the analysis of the discovered trends, the use of SOM techniques is advocated so that similar trends can be grouped together. To identify “interesting” trends a sequences of SOMs are generated which can be interpreted by considering how trends move from one SOM to the next. The further a trend moves from one SOM to the next, the more “interesting” the trend is deemed to be. The study is focused two types of network, Star networks and Complex star networks, exemplified by two real applications: the Cattle Tracing System in operation in Great Britain and a car insurance quotation application. [Copyright &y& Elsevier]

Abstract: Complex variability is a significant problem in predicting construction crew productivity. Neural Networks using supervised learning methods like Feed Forward Back Propagation (FFBP) and General Regression Neural Networks (GRNN) have proved to be more advantageous than statistical methods like multiple regression, considering factors like the modelling ease and the prediction accuracy. Neural Networks using unsupervised learning like Self Organizing Maps (SOM) have additionally been introduced as methods that overcome some of the weaknesses of supervised learning methods through their clustering ability. The objective of this article is thus to compare the performances of FFBP, GRNN and SOM in predicting construction crew productivity. Related data has been collected from 117 plastering crews through a systematized time study and comparison of prediction performances of the three methods showed that SOM have a superior performance in predicting plastering crew productivity. [Copyright &y& Elsevier]

Abstract: In this paper, an adaptive neural system has been built to predict the risk of disruption at ASDEX Upgrade. The system contains a Self Organizing Map, which determines the ‘novelty’ of the input of a Multi Layer Perceptron predictor module. The answer of the MLP predictor will be inhibited whenever a novel sample is detected. Furthermore, it is possible that the predictor produces a wrong answer although it is fed with known samples. In this case, a retraining procedure will be performed to update the MLP predictor in an incremental fashion using data coming from both the novelty detection, and from wrong predictions. In particular, a new update is performed whenever a missed alarm is triggered by the predictor. The performance of the adaptive predictor during the more recent experimental campaigns until November 2009 has been evaluated. [Copyright &y& Elsevier]

Abstract: The state-of-the-art text clustering methods suffer from the huge size of documents with high-dimensional features. In this paper, we studied fast SOM clustering technology for Text Information. Our focus is on how to enhance the efficiency of text clustering system whereas high clustering qualities are also kept. To achieve this goal, we separate the system into two stages: offline and online. In order to make text clustering system more efficient, feature extraction and semantic quantization are done offline. Although neurons are represented as numerical vectors in high-dimension space, documents are represented as collections of some important keywords, which is different from many related works, thus the requirement for both time and space in the offline stage can be alleviated. Based on this scenario, fast clustering techniques for online stage are proposed including how to project documents onto output layers in SOM, fast similarity computation method and the scheme of Incremental clustering technology for real-time processing, We tested the system using different datasets, the practical performance demonstrate that our approach has been shown to be much superior in clustering efficiency whereas the clustering quality are comparable to traditional methods. [Copyright &y& Elsevier]

Short term electricity load forecasting is nowadays of paramount importance in order to estimate next day load, resulting in energy save and environment protection. Electricity demand is influenced (among other things) by the day of the week, the time of year and special periods and/or days, such as religious and national events all of which must be identified prior to modeling. This identification, known as day-type identification, must be included in the modeling stage either by segmenting the data and modeling each day-type separately or by including the day-type as an input. In this study, a two stage clustering approach, based on unsupervised clustering methods is examined to identify regional Algerian electricity load day types. For instance, SOM (Self Organizing Maps) and fuzzy C-means will be presented and applied in detail to load day-type identification with the investigation of the fuzzy C-means clustering algorithm. The optimal number of clusters is obtained using fuzzy cluster validation measures, such as Xie and Beni's index (XB), among others. This application shows that the two-level clustering approach, can effectively identify load day types, for subsequent prediction modeling stages given a possible weighted classification among clusters for particular days. [ABSTRACT FROM AUTHOR]

Variable selection is an important task in machine learning and data mining applications. In many real world problems a huge volume of data is often available which corresponds to a large number of variables. When developing a model for classification, clustering or other applications, the search of an optimal subset of relevant input variables is crucial. In this paper an automatic variable selection method, which combines Genetic Algorithms and Self Organizing Maps, is proposed for classification purposes. The Genetic Algorithm is used to select the most relevant input variables and to set some relevant parameters of a classifier implemented through a Self Organizing Map. This method has been tested with several datasets belonging to the UCI repository. The results of the tests are presented and discussed in this paper. The proposed approach provides a good classification accuracy and contributes to the comprehension of the phenomenon under consideration. [ABSTRACT FROM AUTHOR]

Abstract: A Self Organizing Map (SOM), is a machine learning method that represents high-dimensional data in low-dimensional form without losing topological relations of the data. After an unsupervised learning process, it organizes the data on the basis on similarity. In the current study, a SOM based algorithm has been developed which not only produces 2-D maps to analyze the relationship between various factors and crew productivity, but also predicts productivity under given conditions. Validation of the model has been achieved both by using artificial data set and data from 144 concrete pouring, 101 formwork and 101 reinforcement crews. The results show that maps which are produced by the model are satisfactory in clustering the data and prediction performance of the model is superior to similar artificial neural network models. [Copyright &y& Elsevier]

Abstract: Broadcast-based DSM multiprocessors are nowadays an attractive platform for parallel computing due to their advantages in terms of scalability and programmability. In order to obtain high performance out of these systems, network latency reduction techniques should be developed, which requires the knowledge of the relationship between latency and other important DSM parameters. In this paper, self organizing maps (SOM) are used to investigate the effect of DSM parameters on network latency for a multiprocessor architecture interconnected by the Simultaneous Optical Multiprocessor Exchange Bus (SOME-Bus). An event-based discrete simulator using OPNET Modeler is developed for simulating a SOME-Bus system containing 64-nodes. Two thousand data points have been collected in order to create the dataset used in this study. 2-dimensional (2D) maps are produced by using SOM to display the relationship between network latency and other parameters such as the miss rate to a modified block (P m), fraction of write misses (P w), probability of having an upgrade ownership request message (P uor), probability of having a cache full (P cf) and ratio of the mean thread run time to mean message transfer time (T/R). The results show that the most dominant DSM parameter effecting network latency is T/R, while P cf has the minimum affect on network latency among other parameters. The individual effect of P m, P w and P uor on network latency depends on the values of all other parameters. [Copyright &y& Elsevier]

Abstract: An increasing number of computational and statistical approaches have been used for text classification, including nearest-neighbor classification, naïve Bayes classification, support vector machines, decision tree induction, rule induction, and artificial neural networks. Among these approaches, naïve Bayes classifiers have been widely used because of its simplicity. Due to the simplicity of the Bayes formula, the naïve Bayes classification algorithm requires a relatively small number of training data and shorter time in both the training and classification stages as compared to other classifiers. However, a major short coming of this technique is the fact that the classifier will pick the highest probability category as the one to which the document is annotated too. Doing this is tantamount to classifying using only one dimension of a multi-dimensional data set. The main aim of this work is to utilize the strengths of the self organizing map (SOM) to overcome the inadvertent dimensionality reduction resulting from using only the Bayes formula to classify. Combining the hybrid system with new ranking techniques further improves the performance of the proposed document classification approach. This work describes the implementation of an enhanced hybrid classification approach which affords a better classification accuracy through the utilization of two familiar algorithms, the naïve Bayes classification algorithm which is used to vectorize the document using a probability distribution and the self organizing map (SOM) clustering algorithm which is used as the multi-dimensional unsupervised classifier. [Copyright &y& Elsevier]

Abstract: In this paper, a generalized application of Kohonen Network for automatic point correspondence of unimodal medical images is presented. Given a pair of two-dimensional medical images of the same anatomical region and a set of interest points in one of the images, the algorithm detects effectively the set of corresponding points in the second image, by exploiting the properties of the Kohonen self organizing maps (SOMs) and embedding them in a stochastic optimization framework. The correspondences are established by determining the parameters of local transformations that map the interest points of the first image to their corresponding points in the second image. The parameters of each transformation are computed in an iterative way, using a modification of the competitive learning, as implemented by SOMs. The proposed algorithm was tested on medical imaging data from three different modalities (CT, MR and red-free retinal images) subject to known and unknown transformations. The quantitative results in all cases exhibited sub-pixel accuracy. The algorithm also proved to work efficiently in the case of noise corrupted data. Finally, in comparison to a previously published algorithm that was also based on SOMs, as well as two widely used techniques for detection of point correspondences (template matching and iterative closest point), the proposed algorithm exhibits an improved performance in terms of accuracy and robustness. [Copyright &y& Elsevier]

The applicability of artificial neural networks (ANN) for modelling of daily river flows in a humid tropical river basin with seasonal rainfall pattern is investigated and the model performance assessed using the commonly adopted efficiency indices. Although the developed model showed satisfactory results for rainy period, the predicted hydrograph for the low flow period deviate from the observed data considerably. The rainfall and discharge data available for modelling is explored using Self Organizing Maps (SOM) and the subset of data having definite relationship between the selected hydrologic variables identified. The alternate approach for modelling of river flows utilising the knowledge from SOM analysis has improved the model results. The results show that ANN models can be adopted for forecasting of river flows in the humid tropical river basins for the monsoon period. Input data exploration using SOM is found helpful for developing logically sound ANN models. [ABSTRACT FROM AUTHOR]

Abstract: The significant growth of consumer credit has resulted in a wide range of statistical and non-statistical methods for classifying applicants in ‘good’ and ‘bad’ risk categories. Self organizing maps (SOMs) exist since decades and although they have been used in various application areas, only little research has been done to investigate their appropriateness for credit scoring. This is mainly due to the unsupervised character of the SOM''s learning process. In this paper, the potential of SOMs for credit scoring is investigated. First, the powerful visualization capabilities of SOMs for exploratory data analysis are discussed. Afterwards, it is shown how a trained SOM can be used for classification and how the basic SOM-algorithm can be integrated with supervised techniques like the multi-layered perceptron. Two different methods of integration are proposed. The first technique consists of improving the predictive power of individual neurons of the SOM with the aid of supervised classifiers. The second integration method is similar to a stacking model in which the output of a supervised classifier is entered as an input variable for the SOM. Classification accuracy of both approaches is benchmarked with results reported previously. [Copyright &y& Elsevier]

Abstract: Understanding relationships in high-dimension datasets requires proper data visualization. Two examples of high-dimension data are major-element geochemical and hyperspectral data. Major-element geochemical data consists of eleven oxide measurements for each sample. Well-known correlations exist for these types of data, i.e., the negative relationship between SiO2 and MgO; other more subtle relationships are rarely apparent. Hyperspectral data is by definition high-dimension data consisting of upwards of 100+ discrete measurements of the electromagnetic spectrum for a material. Hyperspectral data are a significant challenge to interpret when evaluating information for heterogeneous materials such as rocks. Self-organizing maps (SOMs) provide insight into complex relationships in high-dimension datasets while preserving the inherent topological relations and simultaneously producing a statistical model of the dataset. Another benefit of SOMs is their generation of composite vectors which can be analyzed to extract the relative importance of each component during classification. The veracity of SOMs is demonstrated using two datasets from the Spanish peaks intrusive complex of south-central Colorado including major-element geochemical and hyperspectral measurements. [Copyright &y& Elsevier]

A hierarchical neural-network-based approach for circuit tuning at the post-fabrication stage is proposed. In this approach,measurements that characterize the behavior of the circuit under test are first selected. The best candidates of circuit parameters for tuning are also determined. A training set comprising the selected circuit measurements is then constructed. These measurements are calculated during simulations in which the circuit parameter values are uniformly distributed in a tolerance region around their nominal values. The training set is fed to a self organizing map neural network to cluster the measurements. The generated clusters are manipulated and classified via a hierarchical circuit tuning procedure. Based on this classification, tuning values for the tuning parameters are calculated. Situations in which the circuit cannot be tuned are also addressed. Experimental results indicate that the developed approach provides a robust and efficient technique for circuit tuning. [ABSTRACT FROM AUTHOR]

• A new function for computing local self-organizing clustering center with SOM clustering algorithm. • An adaptive sign function is presented to replace the constant of the area term in the level set method that enables model to change motion direction of the level set. • An improved double-well potential function is proposed to improve the stability of curve evolution. • The proposed model shows better segmentation accuracy and robustness. Supervised active contour models can use information extracted from supervised samples to guide contour evolution. However, their applicability is limited by the accuracy of the probabilistic models they use, especially when processing images with intensity inhomogeneity. In this paper, an unsupervised activity contour model with Self Organizing Maps (SOM) is proposed. The proposed model employs the self-organizing neural network to perform clustering calculation and the clustering center is called local self-organizing clustering center. An adaptive sign function is used to control the direction of curve evolution. To improve the stability of curve evolution, an improved double-well potential function is proposed. The experiment results show that our model can effectively segment images with intensity inhomogeneity. [ABSTRACT FROM AUTHOR]

[Display omitted] The aim of this study was to apply artificial neural networks as deep learning tools in establishing a model for understanding and prediction of diazepam release from fused deposition modeling (FDM) printed tablets. Diazepam printed tablets of various shapes were created by a computer-aided design (CAD) program and prepared by fused deposition modeling using previously prepared polyvinyl alcohol/diazepam filaments via hot-melt extrusion. The surface to volume ratio (SA/V) for each shape was calculated. Printing parameters were varied including infill density (20%, 70% and 100%) and infill pattern (line and zigzag). Influence of tablet SA/V ratio and printing parameters (infill density and infill pattern) on the release of diazepam from printed tablets were modeled using self-organizing maps (SOM) and multi-layer perceptron (MLP). SOM as an unsupervised neural network was used for visualizing interrelation among the data, whereas MLP was used for the prediction of drug release properties. MLP had three layers (with structure 2-3-5) and was trained using back propagation algorithm. Input parameters for the modeling were: infill density and SA/V ratio; while output parameters were percent of drug release in five time points. The data set for network training was divided into training, validation and test sets. The dissolution rate increased with higher SA/V ratio, lower infill density (less than 50%) and zigzag infill pattern. The established ANN model was tested; calculated f 2 factors for two tested formulations (70.24 and 77.44) showed similarity between experimentally observed and predicted drug release profiles. Trained MLP network was able to predict drug release behavior as a function of infill density and SA/Vol ratio, as established design space for formulated 3D printed diazepam tablets. [ABSTRACT FROM AUTHOR]

Observing the vertical dynamic of phytoplankton in the water column is essential to understand the evolution of the ocean primary productivity under climate change and the efficiency of the CO 2 biological pump. This is usually made through in-situ measurements. In this paper, we propose a machine learning methodology to infer the vertical distribution of phytoplankton pigments from surface satellite observations, allowing their global estimation with a high spatial and temporal resolution. After imputing missing values through iterative completion Self-Organizing Maps, smoothing and reducing the vertical distributions through principal component analysis, we used a Self-Organizing Map to cluster the reduced profiles with satellite observations. These referent vector clusters were then used to invert the vertical profiles of phytoplankton pigments. The methodology was trained and validated on the MAREDAT dataset and tested on the Tara Oceans dataset. The different regression coefficients R 2 between observed and estimated vertical profiles of pigment concentration are, on average, greater than 0.7. We could expect to monitor the vertical distribution of phytoplankton types in the global ocean. [ABSTRACT FROM AUTHOR]