Ivan Jordanov, Nedyalko Petrov and Alessio Petrozziello
In this paper we investigate further and extend our previous work on radar signal identification and classification based on a data set which comprises continuous, discrete and categorical data that represent radar pulse train characteristics such as signal frequencies, pulse repetition, type of modulation, intervals, scan period, scanning type, etc. As the most of the real world datasets, it also contains high percentage of missing values and to deal with this problem we investigate three imputation techniques: Multiple Imputation (MI); K-Nearest Neighbour Imputation (KNNI); and Bagged Tree Imputation (BTI). We apply these methods to data samples with up to 60% missingness, this way doubling the number of instances with complete values in the resulting dataset. The imputation models performance is assessed with Wilcoxon’s test for statistical significance and Cohen’s effect size metrics. To solve the classification task, we employ three intelligent approaches: Neural Networks (NN); Support Vector Machines (SVM); and Random Forests (RF). Subsequently, we critically analyse which imputation method influences most the classifiers’ performance, using a multiclass classification accuracy metric, based on the area under the ROC curves. We consider two superclasses (‘military’ and ‘civil’), each containing several ‘subclasses’, and introduce and propose two new metrics: inner class accuracy (IA); and outer class accuracy (OA), in addition to the overall classification accuracy (OCA) metric. We conclude that they can be used as complementary to the OCA when choosing the best classifier for the problem at hand.
In this paper, we describe how several soft computing tools can be used to assist in high throughput screening of potential drug candidates. Individual small molecules (ligands) are assessed for their potential to bind to specific proteins (receptors). Committees of multilayer networks are used to classify protein-ligand complexes as good binders or bad binders, based on selected chemical descriptors. The novel aspects of this paper include the use of statistical analyses on the weights of single layer networks to select the appropriate descriptors, the use of Monte Carlo cross-validation to provide confidence measures of network performance (and also to identify problems in the data), the addition of new chemical descriptors to improve network accuracy, and the use of Self Organizing Maps to analyze the performance of the trained network and identify anomalies. We demonstrate the procedures on a large practical data set, and use them to discover a promising characteristic of the data. We also perform virtual screenings with the trained networks on a number of benchmark sets and analyze the results.
Recent demand on the measurement resolution of precise positioning comes up to tens of picometers. Some distinguished researches have been performed to measure the displacement in picometer order, however, few of them can verify the measurement performance as available tools in industry. This is not only because the picometer displacement is not yet required for industrial use, but also due to the lack of standard tools to verify such precise displacement. We proposed a displacement reduction mechanism for generating precise displacement using torsional leaf spring hinges (TLSHs) that consist of four leaf springs arranged radially. It has been demonstrated that a prototype of the reduction mechanism was able to provide one-nanometer displacement with 1/1000 reduction rate by a piezoelectric actuator. In order to clarify the potential of the reduction mechanism, a displacement reduction table that can be mounted on AFM stage was newly developed using TLSHs. This paper describes the design of the reduction mechanism and the sub-nanometer displacement performance of the table obtained from its dynamic and static characteristics measured by displacement sensors and from the AFM images
Cristian Rodrìguez Rivero, Juliàn Pucheta, Sergio Laboret, Vìctor Sauchelli and Daniel Patiǹo
This article presents short-term predictions using neural networks tuned by energy associated to series based-predictor filter for complete and incomplete datasets. A benchmark of high roughness time series from Mackay Glass (MG), Logistic (LOG), Henon (HEN) and some univariate series chosen from NN3 Forecasting Competition are used. An average smoothing technique is assumed to complete the data missing in the dataset. The Hurst parameter estimated through wavelets is used to estimate the roughness of the real and forecasted series. The validation and horizon of the time series is presented by the 15 values ahead. The performance of the proposed filter shows that even a short dataset is incomplete, besides a linear smoothing technique employed; the prediction is almost fair by means of SMAPE index. Although the major result shows that the predictor system based on energy associated to series has an optimal performance from several chaotic time series, in particular, this method among other provides a good estimation when the short-term series are taken from one point observations.
Combinatorial optimization problems, such as travel salesman problem, are usually NP-hard and the solution space of this problem is very large. Therefore the set of feasible solutions cannot be evaluated one by one. The simple genetic algorithm is one of the most used evolutionary computation algorithms, that give a good solution for TSP, however, it takes much computational time. In this paper, Affinity Propagation Clustering Technique (AP) is used to optimize the performance of the Genetic Algorithm (GA) for solving TSP. The core idea, which is clustering cities into smaller clusters and solving each cluster using GA separately, thus the access to the optimal solution will be in less computational time. Numerical experiments show that the proposed algorithm can give a good results for TSP problem more than the simple GA.
The present paper1 aims to propose a new type of information-theoretic method to maximize mutual information between inputs and outputs. The importance of mutual information in neural networks is well known, but the actual implementation of mutual information maximization has been quite difficult to undertake. In addition, mutual information has not extensively been used in neural networks, meaning that its applicability is very limited. To overcome the shortcoming of mutual information maximization, we present it here in a very simplified manner by supposing that mutual information is already maximized before learning, or at least at the beginning of learning. The method was applied to three data sets (crab data set, wholesale data set, and human resources data set) and examined in terms of generalization performance and connection weights. The results showed that by disentangling connection weights, maximizing mutual information made it possible to explicitly interpret the relations between inputs and outputs.
Estimating the fundamental frequency and harmonic parameters is basic for signal modeling in a power supply system. This paper presents a complexity-reduced algorithm for signal reconstruction in the time domain from irregularly spaced sampling values. Differing from the existing parameter estimation algorithms, either in power quality monitoring or in harmonic compensation, the proposed algorithm enables a simultaneous estimation of the fundamental frequency, the amplitudes and phases of harmonic waves. The reduction in complexity is achieved owing to completely new analytical and summarized expressions that enable a quick estimation at a low numerical error. It is proved that the estimation performance of the proposed algorithm can attain Cramer-Rao lower bound (CRLB) for sufficiently high signal-to-noise ratios. The proposed algorithm can be applied in signal reconstruction, spectral estimation, system identification, as well as in other important signal processing problems. The simulation and experimental results verify the effectiveness of the proposed algorithm.
Ant colony optimization (stocktickerACO) is a meta-heuristic algorithm inspired by food searching behavior of real ants. Recently stocktickerACO has been widely used in digital image processing. When artificial ants move in a discrete habitat like an image, they deposit pheromone in their prior position. Simultaneously, vaporizing of pheromone in each iteration step avoids from falling in the local minima trap. Iris recognition because of its great dependability and non-invasion has various applications. simulation results demonstrate stocktickerACO algorithm can effectively extract the iris texture. Also it is not sensitive to nuisance factors. Moreover, stocktickerACO in this research preserves details of the various synthetic and real images. Performance of ACO in iris segmentation is compared with operation of traditional approaches such as canny, robert, and sobel edge detections. Experimental results reveal high quality and quite promising of stocktickerACO to segment images with irregular and complex structures.
Clustering is widely used to explore and understand large collections of data. K-means clustering method is one of the most popular approaches due to its ease of use and simplicity to implement. This paper introduces Density-based Split- and -Merge K-means clustering Algorithm (DSMK-means), which is developed to address stability problems of standard K-means clustering algorithm, and to improve the performance of clustering when dealing with datasets that contain clusters with different complex shapes and noise or outliers. Based on a set of many experiments, this paper concluded that developed algorithms “DSMK-means” are more capable of finding high accuracy results compared with other algorithms especially as they can process datasets containing clusters with different shapes, densities, or those with outliers and noise.
Jun-Bao Li, Jing Liu, Jeng-Shyang Pan and Hongxun Yao
Magnetic Resonance Super-resolution Imaging Measurement (MRIM) is an effective way of measuring materials. MRIM has wide applications in physics, chemistry, biology, geology, medical and material science, especially in medical diagnosis. It is feasible to improve the resolution of MR imaging through increasing radiation intensity, but the high radiation intensity and the longtime of magnetic field harm the human body. Thus, in the practical applications the resolution of hardware imaging reaches the limitation of resolution. Software-based super-resolution technology is effective to improve the resolution of image. This work proposes a framework of dictionary-optimized sparse learning based MR super-resolution method. The framework is to solve the problem of sample selection for dictionary learning of sparse reconstruction. The textural complexity-based image quality representation is proposed to choose the optimal samples for dictionary learning. Comprehensive experiments show that the dictionary-optimized sparse learning improves the performance of sparse representation.