Simple Dataset Research Articles

MODIFIED SELECTION OF INITIAL CENTROIDS FOR K- MEANS ALGORITHM

This study focuses on the improved initialization of initial centroids instead of random selection for the K-means algorithm. The random selection of initial seeds is a major drawback of the original Kmeans algorithm because it leads to less reliable result of clustering the data. The modified approach of the k-means algorithm integrates the computation of the weighted mean to improve the seeds initialization. This paper shows the comparison of K-Means and Modified K-Means algorithm, using the first simple dataset of four objects and the dataset for service vehicles. The two simple applications proved that the Modified K- Means of selecting initial centroids is more reliable than K-Means Algorithm. Clustering is better achieved in the modified k-means algorithm. Article DOI: http://dx.doi.org/10.20319/mijst.2016.22.4864 This work is licensed under the Creative Commons Attribution-Non-commercial 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.

Online Learning Algorithm Based on Adaptive Control Theory.

This paper proposes a new online learning algorithm which is based on adaptive control (AC) theory, thus, we call this proposed algorithm as AC algorithm. Comparing to the gradient descent (GD) and exponential gradient (EG) algorithm which have been applied to online prediction problems, we find a new form of AC theory for online prediction problems and investigate two key questions: how to get a new update law which has a tighter upper bound on the error than the square loss? How to compare the upper bound for accumulated losses for the three algorithms? We obtain a new update law which fully utilizes model reference AC theory. Moreover, we present upper bound on the worst-case expected loss for AC algorithm and compare it with previously known bounds for the GD and EG algorithm. The loss bound we get in this paper is a time-varying function, which provides increasingly accurate estimates for upper bound. The AC algorithm has a much smaller loss only if the number of the samples meets certain conditions which can be seen in this paper. We also performed experiments which show that our update law is reasonably feasible and our upper bound is quite tight on both simple artificial and real data sets. The main contributions of this paper are twofold. First of all, we develop a new online algorithm called AC algorithm, and second, we obtain improved bounds, see Theorems 2-4 in this paper.

Two dimensional smoothing via an optimised Whittaker smoother

BackgroundIn many applications where moderate to large datasets are used, plotting relationships between pairs of variables can be problematic. A large number of observations will produce a scatter-plot which is difficult to investigate due to a high concentration of points on a simple graph.In this article we review the Whittaker smoother for enhancing scatter-plots and smoothing data in two dimensions. To optimise the behaviour of the smoother an algorithm is introduced, which is easy to programme and computationally efficient.ResultsThe methods are illustrated using a simple dataset and simulations in two dimensions. Additionally, a noisy mammography is analysed. When smoothing scatterplots the Whittaker smoother is a valuable tool that produces enhanced images that are not distorted by the large number of points. The methods is also useful for sharpening patterns or removing noise in distorted images.ConclusionThe Whittaker smoother can be a valuable tool in producing better visualisations of big data or filter distorted images. The suggested optimisation method is easy to programme and can be applied with low computational cost.

Appropriate Number of Standard 2 × 2 Max Pooling Layers and Their Allocation in Convolutional Neural Networks for Diverse and Heterogeneous Datasets

A problem of appropriately allocating pooling layers in convolutional neural networks is considered. The consideration is based on CIFAR-10, NORB, and EEACL26 datasets for preventing “overfitting” in a solution of the problem. For highly accurate image recognition within these datasets, the networks are used with the max pooling operation. The most common form of such operation, which is a 2 ´ 2 pooling layer, is applied with a stride of 2 without padding after convolutional layers. Based on performance against a series of the network architectures, a rule for the best allocation of max pooling layers is formulated. The rule is to insert a few pooling layers after the starting convolutional layers and to insert a one pooling layer after the last but one convolutional layer (“11...100...010”). For much simpler datasets, the best allocation is “11...100...0”.

Team Member Perceptions of Alternative Decision Analysis Approaches

This research evaluated team members’ perceptions of alternative decision analysis approaches when they were asked to analyze a simple or complex data set under a time constraint. There has been significant work done to understand what criteria should be included in this class of problems. However, less work has been done examining the selection of a specific decision analysis approach. This is an area of concern since research has shown that selecting a different decision analysis approach can change the results even when the same data is analyzed. This research effort is a preliminary look at perceptions of alternative decision analysis approaches. A mixed method study analyzing quantitative and qualitative data was completed. Teams were assigned to use one of five different decision analysis approaches, including Weighted Sum, SMART, TOPSIS, AHP, and ELECTRE I. After applying the approach to either a simple or complex data set, team members were surveyed to determine perceptions including comfort with the approach, its results, and ease/difficulty of explaining the approach to coworkers. Discussions were also held with team members to gain additional insights. The study indicated that team members preferred less technically challenging decision analysis approaches, such as Weighted Sum and to a lesser degree SMART, when they were using a simple data set because they found the approach easier to use and to explain to others. The more technically challenging decision analysis approaches, such as AHP and ELECTRE I, were preferred with a complex data set because they provided better insight into the problem. Teams in this study did selected different alternatives when they used different decision analysis approaches. Demographic data was also collected but was not found to influence results in this study.

Ship Detection and Classification on Optical Remote Sensing Images Using Deep Learning

Ship detection and classification is critical for national maritime security and national defense. Although some SAR (Synthetic Aperture Radar) image-based ship detection approaches have been proposed and used, they are not able to satisfy the requirement of real-world applications as the number of SAR sensors is limited, the resolution is low, and the revisit cycle is long. As massive optical remote sensing images of high resolution are available, ship detection and classification on theses images is becoming a promising technique, and has attracted great attention on applications including maritime security and traffic control. Some digital image processing methods have been proposed to detect ships in optical remote sensing images, but most of them face difficulty in terms of accuracy, performance and complexity. Recently, an autoencoder-based deep neural network with extreme learning machine was proposed, but it cannot meet the requirement of real-world applications as it only works with simple and small-scaled data sets. Therefore, in this paper, we propose a novel ship detection and classification approach which utilizes deep convolutional neural network (CNN) as the ship classifier. The performance of our proposed ship detection and classification approach was evaluated on a set of images downloaded from Google Earth at the resolution 0.5m. 99% detection accuracy and 95% classification accuracy were achieved. In model training, 75× speedup is achieved on 1 Nvidia Titanx GPU.

Machine-learned cluster identification in high-dimensional data

BackgroundHigh-dimensional biomedical data are frequently clustered to identify subgroup structures pointing at distinct disease subtypes. It is crucial that the used cluster algorithm works correctly. However, by imposing a predefined shape on the clusters, classical algorithms occasionally suggest a cluster structure in homogenously distributed data or assign data points to incorrect clusters. We analyzed whether this can be avoided by using emergent self-organizing feature maps (ESOM). MethodsData sets with different degrees of complexity were submitted to ESOM analysis with large numbers of neurons, using an interactive R-based bioinformatics tool. On top of the trained ESOM the distance structure in the high dimensional feature space was visualized in the form of a so-called U-matrix. Clustering results were compared with those provided by classical common cluster algorithms including single linkage, Ward and k-means. ResultsWard clustering imposed cluster structures on cluster-less “golf ball”, “cuboid” and “S-shaped” data sets that contained no structure at all (random data). Ward clustering also imposed structures on permuted real world data sets. By contrast, the ESOM/U-matrix approach correctly found that these data contain no cluster structure. However, ESOM/U-matrix was correct in identifying clusters in biomedical data truly containing subgroups. It was always correct in cluster structure identification in further canonical artificial data. Using intentionally simple data sets, it is shown that popular clustering algorithms typically used for biomedical data sets may fail to cluster data correctly, suggesting that they are also likely to perform erroneously on high dimensional biomedical data. ConclusionsThe present analyses emphasized that generally established classical hierarchical clustering algorithms carry a considerable tendency to produce erroneous results. By contrast, unsupervised machine-learned analysis of cluster structures, applied using the ESOM/U-matrix method, is a viable, unbiased method to identify true clusters in the high-dimensional space of complex data.

A methodology for the design of experiments in computational intelligence with multiple regression models.

The design of experiments and the validation of the results achieved with them are vital in any research study. This paper focuses on the use of different Machine Learning approaches for regression tasks in the field of Computational Intelligence and especially on a correct comparison between the different results provided for different methods, as those techniques are complex systems that require further study to be fully understood. A methodology commonly accepted in Computational intelligence is implemented in an R package called RRegrs. This package includes ten simple and complex regression models to carry out predictive modeling using Machine Learning and well-known regression algorithms. The framework for experimental design presented herein is evaluated and validated against RRegrs. Our results are different for three out of five state-of-the-art simple datasets and it can be stated that the selection of the best model according to our proposal is statistically significant and relevant. It is of relevance to use a statistical approach to indicate whether the differences are statistically significant using this kind of algorithms. Furthermore, our results with three real complex datasets report different best models than with the previously published methodology. Our final goal is to provide a complete methodology for the use of different steps in order to compare the results obtained in Computational Intelligence problems, as well as from other fields, such as for bioinformatics, cheminformatics, etc., given that our proposal is open and modifiable.

Regularized forecasting of chaotic dynamical systems

While local models of dynamical systems have been highly successful in terms of using extensive data sets observing even a chaotic dynamical system to produce useful forecasts, there is a typical problem as follows. Specifically, with k-near neighbors, kNN method, local observations occur due to recurrences in a chaotic system, and this allows for local models to be built by regression to low dimensional polynomial approximations of the underlying system estimating a Taylor series. This has been a popular approach, particularly in context of scalar data observations which have been represented by time-delay embedding methods. However such local models can generally allow for spatial discontinuities of forecasts when considered globally, meaning jumps in predictions because the collected near neighbors vary from point to point. The source of these discontinuities is generally that the set of near neighbors varies discontinuously with respect to the position of the sample point, and so therefore does the model built from the near neighbors. It is possible to utilize local information inferred from near neighbors as usual but at the same time to impose a degree of regularity on a global scale. We present here a new global perspective extending the general local modeling concept. In so doing, then we proceed to show how this perspective allows us to impose prior presumed regularity into the model, by involving the Tikhonov regularity theory, since this classic perspective of optimization in ill-posed problems naturally balances fitting an objective with some prior assumed form of the result, such as continuity or derivative regularity for example. This all reduces to matrix manipulations which we demonstrate on a simple data set, with the implication that it may find much broader context.

Descriptors and regions of interest fusion for in- and cross-database gender classification in the wild

Gender classification (GC) has achieved high accuracy in different experimental evaluations based mostly on inner facial details. However, these results do not generalize well in unrestricted datasets and particularly in cross-database experiments, where the performance drops drastically. In this paper, we analyze the state-of-the-art GC accuracy on three large datasets: MORPH, LFW and GROUPS. We discuss their respective difficulties and bias, concluding that the most challenging and wildest complexity is present in GROUPS. This dataset covers hard conditions such as low resolution imagery and cluttered background. Firstly, we analyze in depth the performance of different descriptors extracted from the face and its local context on this dataset. Selecting the bests and studying their most suitable combination allows us to design a solution that beats any previously published results for GROUPS with the Dago's protocol, reaching an accuracy over 94.2%, reducing the gap with other simpler datasets. The chosen solution based on local descriptors is later evaluated in a cross-database scenario with the three mentioned datasets, and full dataset 5-fold cross validation. The achieved results are compared with a Convolutional Neural Network approach, achieving rather similar marks. Finally, a solution is proposed combining both focuses, exhibiting great complementarity, boosting GC performance to beat previously published results in GC both cross-database, and full in-database evaluations.

Development of LiDAR Database Management System using Open Source Software

This study is focused on the development of a LiDAR database management system for the Department of Survey and Mapping Malaysia (JUPEM) to facilitate user authentication, retrieval of LiDAR datasets, storage, scheduling of LiDAR flight sessions and generation of new data products. Additionally, the design goal of the data management system is to support managers-vendors relationship, as well as new data generation out of the results. In this study, we described the structure development of LiDAR database management system and how the system collaborated with data production and data acquisition. The architecture of such application in WebGIS, providing map interactivity in displaying a simple dataset of LiDAR data by using Open Layers integration via GeoJSON format as the spatial data were used to show the implementation of such feature. We also used the Open Source Software (OSS) in the development of the LiDAR Database Management System for JUPEM. This was because many countries, especially in the public sector, were slowly transferring from using proprietary software to OSS.

Comparative study on biophysical ecosystem service mapping methods—a test case of carbon stocks in Finnish Forest Lapland

Mapping of ecosystem services in biophysical environment helps distinguishing the nature’s benefits in our surroundings. Ecosystem services also enable assessing the level of sustainability of the use of the environment. Various methods are being used for mapping them and methods: vary from expert evaluations to complex modelling. All methods: are related with uncertainties and affect to their applicability for different purposes. In this study, two different level biophysical mapping based datasets were tested; 1) data combination of biophysical polygon GIS data (SutiGIS) and statistics, and 2) modelled data of National Forest Inventory were evaluated to compare the differences between the methods: for carbon storages. Our results show that there are differences between results that have been produced in different levels of complexity. However, we also discovered that simpler datasets can also help detecting the shortcomings that may occur in modelling. We discuss also the potential applications of the used datasets in terms of providing information on biodiversity in ES assessments.

Comparison of Atmospheric Pressure Ionization for the Analysis of Heavy Petroleum Fractions with Ion Mobility-Mass Spectrometry

Because of its high molecular complexity, ultra high-resolution Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) is used to characterize heavy petroleum distillates. However, high peak capacity can also be obtained with ion mobility-mass spectrometry (IM-MS). In this case, the additional separation dimension allows for the compensation of the lower resolving power of the time-of-flight mass analyzer. The processing and elemental composition assignments from such bidimensional IM-MS data cannot be achieved with conventional software packages. PetroOrg, a new software package, was used to compare IM-MS data sets recorded with atmospheric solid analysis probe (ASAP) and three other atmospheric pressure ionization sources. The choice of the ionization source is indeed very critical as it defines the type of compounds that can be detected from mixtures. Electrospray ionization (ESI) is very specific to polar compounds and yields relatively simple data sets that can be readily attributed t...

Apex: phylogenetics with multiple genes.

Genetic sequences of multiple genes are becoming increasingly common for a wide range of organisms including viruses, bacteria and eukaryotes. While such data may sometimes be treated as a single locus, in practice, a number of biological and statistical phenomena can lead to phylogenetic incongruence. In such cases, different loci should, at least as a preliminary step, be examined and analysed separately. The r software has become a popular platform for phylogenetics, with several packages implementing distance‐based, parsimony and likelihood‐based phylogenetic reconstruction, and an even greater number of packages implementing phylogenetic comparative methods. Unfortunately, basic data structures and tools for analysing multiple genes have so far been lacking, thereby limiting potential for investigating phylogenetic incongruence. In this study, we introduce the new r package apex to fill this gap. apex implements new object classes, which extend existing standards for storing DNA and amino acid sequences, and provides a number of convenient tools for handling, visualizing and analysing these data. In this study, we introduce the main features of the package and illustrate its functionalities through the analysis of a simple data set.

Complexity curve: a graphical measure of data complexity and classifier performance

We describe a method for assessing data set complexity based on the estimation of the underlining probability distribution and Hellinger distance. In contrast to some popular complexity measures, it is not focused on the shape of a decision boundary in a classification task but on the amount of available data with respect to the attribute structure. Complexity is expressed in terms of graphical plot, which we call complexity curve. It demonstrates the relative increase of available information with the growth of sample size. We perform theoretical and experimental examination of properties of the introduced complexity measure and show its relation to the variance component of classification error. We then compare it with popular data complexity measures on 81 diverse data sets and show that it can contribute to explaining performance of specific classifiers on these sets. We also apply our methodology to a panel of simple benchmark data sets, demonstrating how it can be used in practice to gain insights into data characteristics. Moreover, we show that the complexity curve is an effective tool for reducing the size of the training set (data pruning), allowing to significantly speed up the learning process without compromising classification accuracy. The associated code is available to download at:https://github.com/zubekj/complexity_curve(open source Python implementation).

Assessment of the importance of moderate spatial resolution cloud climatology data in solar radiation modeling

Improving solar radiation models is critical for supporting the increase in solar energy usage and modeling ecosystem dynamics. However, coarse spatial resolutions of solar radiation models overlook the impacts resulting from spatial variability of clouds at meso- and micro-scales. To address this problem, Moderate Resolution Imaging Spectroradiometer (MODIS) cloud climatology developed by the National Severe Storms Laboratory was used to relate cloudiness to surface solar radiation observations. We developed a linear regression model between the surface solar radiation and MODIS cloud climatology and used the model to estimate average radiation across Oklahoma. Furthermore, the study compared the average error and coefficient of determination to measured ground radiation. Error analysis of the regression model showed that the differences between observed radiation and estimated radiation were spatially autocorrelated for the Aqua MODIS satellite scan. This suggests cloudiness alone is not sufficient to predict surface solar radiation. This study found that simple cloud datasets alone can account for approximately 50% of the variation in observed solar radiation at 250-m spatial resolution, but additional datasets such as optical depth, elevation, and slope are needed to accurately explain spatial distributions of incoming shortwave radiation.

Impact of Fiscal Decentralizations on Education and Healthcare Outcomes: Empirical Evidence from Pakistan

This paper aims at investigating the impact of fiscal decentralization on two key social services: education and healthcare outcomes. We use both a simple time series data set and panel data set covering four provinces of Pakistan over the period from 1975 to 2009 to empirically test the impact of fiscal decentralization on infant mortality rate, crude death rate and literacy rate. Besides Ordinary Least Square method we use Generalized Method of Moment econometric technique to obtain robust and consistent results. The empirical findings of this paper indicate that fiscal decentralization is effective in enhancing the delivery and augmenting the quality of education and healthcare services. These findings are important because they suggest, contrary to the traditional public finance theory, that provincial governments can play a far better role in improving the social services delivery like basic education and healthcare compare to federal government.

Impact of Fiscal Decentralizations on Education and Healthcare Outcomes: Empirical Evidence from Pakistan

This paper aims at investigating the impact of fiscal decentralization on two key social services: education and healthcare outcomes. We use both a simple time series data set and panel data set covering four provinces of Pakistan over the period from 1975 to 2009 to empirically test the impact of fiscal decentralization on infant mortality rate, crude death rate and literacy rate. Besides Ordinary Least Square method we use Generalized Method of Moment econometric technique to obtain robust and consistent results. The empirical findings of this paper indicate that fiscal decentralization is effective in enhancing the delivery and augmenting the quality of education and healthcare services. These findings are important because they suggest, contrary to the traditional public finance theory, that provincial governments can play a far better role in improving the social services delivery like basic education and healthcare compare to federal government.

Evaluation and comparison of Genetic Algorithm and Bees Algorithm for location–allocation of earthquake relief centers

One of the important steps in the earthquake disaster management is the establishment of temporary relief centers, to provide the basic helps and support in short time. Finding optimum location for these centers with adequate covering of the urban areas is not a trivial problem. The meta-heuristic algorithms are promising methods, capable of solving such complex optimization problems. The goal of this research was to compare the performance of Genetic Algorithm (GA) and Bees Algorithm (BA) in finding the optimum location of relief centers and in allocating of the parcels to them. In order to limit the search space, GIS was used for selecting a few candidate sites that satisfy the initial conditions and criteria. Then, the two algorithms were used to select nine optimum sites among the candidates and to allocate the parcels to them, while minimizing the sum of all distances between centers and parcels. To calibrate the parameters of the algorithms, a simple simulated data set was used. Having proper values for those parameters, the algorithms were tested on the real data of the study area. The results showed that the convergence of the BA was rather gradual, while the trend for GA was relatively stepwise. Both algorithms showed high levels of repeatability. For both the simulated and real data, GA showed to be faster than BA. Simplicity and repeatability of the algorithm are the main factors from the user’s point of view. Therefore, considering these criteria, the GA is more favorable than the BA.

Introducing Machine Learning Concepts with WEKA.

This chapter presents an introduction to data mining with machine learning. It gives an overview of various types of machine learning, along with some examples. It explains how to download, install, and run the WEKA data mining toolkit on a simple data set, then proceeds to explain how one might approach a bioinformatics problem. Finally, it includes a brief summary of machine learning algorithms for other types of data mining problems, and provides suggestions about where to find additional information.