Abstract
Using machine-learning methods based on self-organising Kohonen maps, the results of numerical simulation of the acceleration of electrons during the interaction of high-power laser ...radiation with plasma are analysed and classified. The particle-in-cell (PIC) method is used to simulate the interaction in a wide range of parameters (laser intensity and plasma concentration). For each set of parameters, the spectrum of accelerated electrons is found, based on which the charge, average energy, and relative energy spread of accelerated electrons are calculated. Using the obtained values as input parameters of the map, the classification of various acceleration regimes is performed. The developed scheme can be used to identify the optimal acceleration regimes under more realistic conditions, considering a larger number of parameters.
Parkinson’s disease is a progressive neurodegenerative condition whose prevalence has significantly increased. This work proposes the development of a severity index to classify patients from ...symptoms, mainly motor ones, using an Artificial Neuronal Network (ANN) trained by the Self-Organizing Maps (SOMs) algorithm. The FOX Insight database was used, which offers data in the form of questionnaires answered by patients or caregivers from all over the world, with information regarding this pathology. After pre-processing the data, a set of 597 questionnaires containing 28 defined questions was selected. The symptoms were individually analyzed after mapping and divided into four classes. In class 1, most symptoms were not present. In class 2, the presence of certain symptoms demonstrated early milestones of the disease. In class 3, symptoms related to the patient’s mobility, in particular pain, stand out among the most reported. In class 4, the intense presence of all symptoms is observed. To test the tool, data were used from some of these patients, who answered the same questionnaire at different times (simulating medical appointments). The presented severity index to classify patients allowed identifying the current stage of the disease allowing the follow-up. This AI-based decision-support tool can help medical professionals to predict the evolution of Parkinson’s disease, which can result in longer life quality of patients, in terms of symptoms and medication requirements.
Solving inverse problems, where we find the input values that result in desired values of outputs, can be challenging. The solution process is often computationally expensive and it can be difficult ...to interpret the solution in high-dimensional input spaces. In this paper, we use a problem from additive manufacturing to address these two issues with the intent of making it easier to solve inverse problems and exploit their results. First, focusing on Gaussian process surrogates that are used to solve inverse problems, we describe how a simple modification to the idea of tapering can substantially speed up the surrogate without losing accuracy in prediction. Unlike block tapering, which approximates the covariance matrix by diagonal blocks, our approach divides the data itself into blocks. Both approaches reduce the computational cost by replacing the Cholesky decomposition of the full matrix by the decomposition of multiple smaller matrices, but our approach gives accurate predictions despite the approximation as we identify hyperparameters optimal for each block. Second, we demonstrate that Kohonen self-organizing maps can be used to visualize and interpret the solution to the inverse problem in the high-dimensional input space. For our data set, as not all input dimensions are equally important, we show that using weighted distances results in a better organized map that not only makes the relationships among the inputs obvious, but also indicates the location of the solution in the input space so an additive manufacturing engineer can control the inputs appropriately for a desired output.
This article is an extended version of a paper presented in the WSOM׳2012 conference (Bourgeois et al., 2012 1). We display a combination of factorial projections, SOM algorithm and graph techniques ...applied to a text mining problem. The corpus contains eight medieval manuscripts which were used to teach arithmetic techniques to merchants.
Among the techniques for Data Analysis, those used for Lexicometry (such as Factorial Analysis) highlight the discrepancies between manuscripts. The reason for this is that they focus on the deviation from the independence between words and manuscripts. Still, we also want to discover and characterize the common vocabulary among the whole corpus.
Using the properties of stochastic Kohonen maps, which define neighborhood between inputs in a non-deterministic way, we highlight the words which seem to play a special role in the vocabulary. We call them fickle and use them to improve both Kohonen map robustness and significance of FCA visualization. Finally we use graph algorithmic to exploit this fickleness for classification of words.
A method of identifying banks’ business models and studying the features of their risk profile, considering the system of indicators featuring the structure of assets, liabilities, income, expenses, ...and other qualitative indicators based on monthly statistical reporting. Kohonen's self-organizing maps (SOM) are used to process large data sets, revealing objects’ hidden features by forming homogeneous groups according to similar values of a large system of indicators. The choice of the system of indicators that play the most significant role in describing the business models of modern banks is substantiated. The proposed method makes it possible to group banks with homogeneous characteristics into so-called structural-functional groups and studies the change in the characteristics of groups of banks over time to compare their behavior during periods of active development of the system and during a crisis. That approach is useful for studying the banking system at the macro level, as it provides a quantitative measure of its financial stability. The more banks are in groups with negative values of parameters, increased risks, and unprofitable performance, the worse the general state of the system. The method also allows studying the features of each structural and functional group and the business models’ features at the meso-level. The number and composition of banks inherent in any group change dynamically, which characterizes the features of the relevant business model in a particular period. The averages of each group reflect the objective changes in the banking system structure. In addition, the SOM trajectory can be built for each individual bank determining the development of its strategy, features of a particular business model, and risk profile. At the micro-level, it allows comparing the features of individual banks within the SFGB and models ways to improve efficiency and financial stability by forecast values for SOM. An extensive system of indicators used to form structural and functional groups of banks allows to quickly respond to changes in the banking system, identify areas of increased risk and explore the adequacy and effectiveness of banks’ business models.
The qualitative analysis of multidimensional data using their visualization allows to observe some characteristics of data in a way which is the most natural for a human, through the sense of sight. ...Thanks to such an approach, some characteristics of the analyzed data are simply visible. This allows to avoid using often complex algorithms allowing to examine specific data properties. Visualization of multidimensional data consists in using the representation transforming a multidimensional space into a two-dimensional space representing a computer screen. The important information which can be obtained in this way is the possibility to separate points belonging to different classes in the multidimensional space. Such information can be directly obtained if images of points belonging to different classes occupy other areas of the picture presenting these data. The paper presents the effectiveness of the qualitative analysis of multidimensional data conducted in this way through their visualization with the application of Kohonen maps and autoassociative neural networks. The obtained results were compared with results obtained using the perspective-based observational tunnels method, PCA, multidimensional scaling and relevance maps. Effectiveness tests of the above methods were performed using real seven-dimensional data describing coal samples in terms of their susceptibility to fluidal gasification. The methods’ effectiveness was compared using the criterion for the readability of the multidimensional visualization results, introduced in earlier papers.
Introduction. Formation of prerequisites for innovative development of industrial enterprises includes strategic planning, which is one of the components of the management system. The choice of the ...strategy of innovative development of the portfolio of strategies for innovation-active enterprises depends on their internal and external environment. Purpose. The aim of the study is to develop conceptual approaches to clustering of innovation-active machine-building enterprises, to analyze the clusters in order to identify their significant differences, and on this basis to develop appropriate strategies for each cluster of enterprises. Results. A conceptual approach to the clustering of innovation-active machine-building enterprises in order to identify the degree of their readiness for innovation. Clustering was carried out for thirty innovation-active enterprises on two indicators – the level of the crisis and the indicator of the use of strategic capabilities of enterprises. By means of the correlation analysis the relative dependence of input factors is established and by its results it is shown that the incoming factors are practically independent from each other. The use of cluster analysis on the basis of modern research methods using neural networks – Kohonen maps allowed to differentiate the innovation-active enterprises according to the degree of their readiness for innovation. The method of neural networks – Kohonen maps was used for clustering. Clustering is carried out on the basis of two input indicators – an indicator of the level of crisis and an indicator of the use of strategic opportunities.These clusters are called «0», «1», «2», «3» and «4». The largest cluster «0» included 13 machine-building innovation-active enterprises, the smallest cluster «1» included one innovation-active enterprise. The enterprises of cluster «3» have average strategic opportunities with moderate negative dynamics of the crisis indicator. In General, this is one of the largest crisis groups of enterprises among the thirty studied innovation-active enterprises. Enterprises of clusters «1» and «2» have average strategic opportunities with stable positive dynamics of the crisis state indicator. These two clusters include only three enterprises out of the thirty innovation-active enterprises under study. Conclusions. All thirty studied enterprises are divided into five clusters, which differ significantly in parameters. For machine-building innovative-active machine-building enterprises of each cluster the corresponding innovative strategy is offered.
The purpose of current investigation is to engage two efficient evolvable neuro-evolutionary machines to identify a nonlinear dynamic model for the shape memory alloy (SMA) actuators. SMA materials ...are kind of smart materials capable of compensating any undergo plastic deformations and return to their memorized shape. This fascinating trait gives them versatility to be applied in different engineering applications such as smart actuators and sensors. As a result, modeling and analyzing of their response is an essential task for researchers. Nevertheless, these materials have intricate behaviors that incorporate the modeling with major dilemma and obstacles. In this research, two novel evolvable machine comprised recurrent neural network (RNN) and two novel hybrid heuristic methods nominally cellular automate and Kohonen map assisted versions of the great salmon run (CTGSR and KTGSR respectively) optimization algorithm are developed to find a robust, representative and reliable recursive identification framework capable of modeling the proposed SMA actuator. To elaborate on the acceptable performance of proposed systems, several experimental tests are carried out. Obtained results reveal the promising potential of the evolvable frameworks in modeling the behavior of SMA as a complex real world engineering system. Furthermore, by executing some comparative tests, authors indicate that both of their proposed hybrid heuristic algorithms outperform the sole version of TGSR as well as some other well-known evolutionary algorithms.
•Applying the evolving computation concept to train a dynamic black box capable of modeling shape memory alloy actuators.•Proposing two novel hybrid systems based on cellular automated and self-organizing maps theory.•Examining the potential of a recent spotlighted metaheuristic called “The Great Salmon Run” for real world applications.•Elaborating on the performance of proposed method in dynamic environment.•Analyzing the behavior of proposed shape memory actuator by using the MSE error at each step.
The purpose of current investigation is to engage two efficient evolvable neuro-evolutionary machines to identify a nonlinear dynamic model for a shape memory alloy (SMA) actuator. SMA materials are kind of smart materials capable of compensating any undergo plastic deformations and return to their memorized shape. This fascinating trait gives them versatility to be applied on different engineering applications such as smart actuators and sensors. As a result, modeling and analyzing of their response is an essential task to researchers. Nevertheless, these materials have intricate behaviors that incorporate the modeling with major dilemma and obstacles. In this research, two novel evolvable machines comprised recurrent neural network (RNN) and two novel hybrid heuristic methods nominally cellular automate and Kohonen map assisted versions of The Great Salmon Run (CTGSR and KTGSR respectively) optimization algorithm are developed to find a robust, representative and reliable recursive identification framework capable of modeling the proposed SMA actuator. To elaborate on the acceptable performance of proposed systems, several experimental tests are carried out. Obtained results reveal the promising potential of the evolvable frameworks for modeling the behavior of SMA as a complex real world engineering system. Furthermore, by executing some comparative tests, the authors indicate that both of their proposed hybrid heuristic algorithms outperform the sole version of TGSR as well as some other well-known evolutionary algorithms.
To interpret and to process the answers to questionnaires with large amount of questions may be not easy task. They are multidimensional data, sometimes with high dimensionality (in the hundreds). ...Therefore, it is necessary that some data reduction approach should be employed. On the other hand, answers to specific questions in questionnaires are imprecise, and the type and degree of imprecision is determined by the kind of the questions. The authors of the paper consider the imprecise answers to management type questions using a numerical scale as fuzzy degrees, and based on the semantic connections among the individual questions, a hierarchical structure is assumed. The paper suggests the use of factor analysis in order to determine this hierarchical structure, and thus the construction of fuzzy signatures from the tree graph representing the connections among the questions and answers, and the values normalized into membership degrees are assigned to the leaves of this tree. An interesting issue is how to determine the aggregations at the intermediate nodes. This may happen based on management science domain expert knowledge, and validated by the obtained results. Kohonen maps are used to demonstrate the clusters emerging among the overall fuzzy degrees representing the Fuzzy Signatures. The evaluation brings some results that partly confirm soft science based assumptions about employee behavior in the literature, and partly bring some interesting novel recognitions that may be brought in feedback to the original management science related problem, where the new method is illustrated.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, UILJ, UKNU, UL, UM, UPUK
Display omitted
► We evaluated effects of SOMs parameters on classification and computational time. ► The study was conducted on eighteen real datasets. ► Significant parameters and their ...interactions on classification were highlighted. ► Optimal architectures to reduce the computational time of SOMs was proposed.
Self Organising Maps (SOMs) are one of the most powerful learning strategies among neural networks algorithms. SOMs have several adaptable parameters and the selection of appropriate network architectures is required in order to make accurate predictions. The major disadvantage of SOMs is probably due to the network optimisation, since this procedure can be often time-expensive.
Effects of network size, training epochs and learning rate on the classification performance of SOMs are known, whereas the effect of other parameters (type of SOMs, weights initialisation, training algorithm, topology and boundary conditions) are not so obvious.
This study was addressed to analyse the effect of SOMs parameters on the network classification performance, as well as on their computational times, taking into consideration a significant number of real datasets, in order to achieve a comprehensive statistical comparison. Parameters were contemporaneously evaluated by means of an approach based on the design of experiments, which enabled the investigation of their interaction effects.
Results highlighted the most important parameters which influence the classification performance and enabled the identification of the optimal settings, as well as the optimal architectures to reduce the computational time of SOMs.