Search

mkFit is an implementation of the Kalman filter-based track reconstruction algorithm that exploits both threadand data-level parallelism. In the past few years the project transitioned from the R&D phase to deployment in the Run-3 offline workflow of the CMS experiment. The CMS tracking performs a series of iterations, targeting reconstruction of tracks of increasing difficulty after removing hits associated to tracks found in previous iterations. mkFit has been adopted for several of the tracking iterations, which contribute to the majority of reconstructed tracks. When tested in the standard conditions for production jobs, speedups in track pattern recognition are on average of the order of 3.5x for the iterations where it is used (3-7x depending on the iteration). Multiple factors contribute to the observed speedups, including vectorization and a lightweight geometry description, as well as improved memory management and single precision. Efficient vectorization is achieved with both the icc and the gcc (default in CMSSW) compilers and relies on a dedicated library for small matrix operations, Matriplex, which has recently been released in a public repository. While the mkFit geometry description already featured levels of abstraction from the actual Phase-1 CMS tracker, several components of the implementations were still tied to that specific geometry. We have further generalized the geometry description and the configuration of the run-time parameters, in order to enable support for the Phase-2 upgraded tracker geometry for the HL-LHC and potentially other detector configurations. The implementation strategy and high-level code changes required for the HL-LHC geometry are presented. Speedups in track building from mkFit imply that track fitting becomes a comparably time consuming step of the tracking chain. Prospects for an mkFit implementation of the track fit are also discussed. [ABSTRACT FROM AUTHOR]

The fragility of long bones in osteoporosis is linked to a decrease in cortical layer thickness (CTh) and the development of internal porosity (P) within the cortex. This study is dedicated to investigate whether it is possible and how accurately P and CTh can be determined using a pattern recognition methods, utilizing discrete Fourier transform (DFT) analysis of ultrasonic signals transmitted through a soft tissue layer. As input data, ultrasound scans of phantom of cortical bone with different grades of CTh and P were used. Each of the phantom scans was chosen in turn as a test object, while other specimens composed the training set. For analysis, three different existing approaches were applied firstly and the later one was developed by improving the existing ones. The analysis was carried out in three categories: P and CTh ratio, P only and CTh only. Based on the results of each analysis, a method was obtained that predicts P and CTh better than other applied ones. A comparison of the methods for P and CTh determining and choosing the most effective ones are shown. [ABSTRACT FROM AUTHOR]

The resin 3D printing is widely recognized as first 3D printing technology and also called as VAT photopolymerization technique, that uses Ultra Violet (UV) laser light and photopolymer (PP) resin to build the component layer by layer. Complex components with fine features, correct measurements, great quality, and high-resolution are features of this technique. In fact, the quality of 3D printing has a lot to do with the equipment, materials and slicer. Although this technology has so many advantages, the dimensional accuracy of printed parts depends on many factors such as layer height, orientation, exposure time. Designers are crucial in conceptualizing a product before it is fabricated, which increases the design's resilience. Manufacturers can use 3D printing to directly build components at this level, but precision must always be maintained. The PP resins are very much prone to the shrinkage after immediate printing and it is not absolutely cured in the machine. It is necessary to do post curing in order to improve structural performance, at the same time the component shrinks, unfortunately, this may result in clearance and fit issues in design perception. Inherent size imperfections are frequent in 3D printed objects, the most common of which being variances in the linear dimension and hole diameter. These two forms of errors were investigated, as well as their effects on the dimensional correctness of a typical part. The experimental results of a preliminary investigation for post-cured specimens printed using the VAT photopolymerization 3D printing technology are presented in this article. The study aimed in predicting the influencing parameters using Taguchi analysis by design of experiments in order to improve the dimensional accuracy (DA) of 3D printed part and a regression equation is developed to predict the better outcomes. Based on the findings of Analysis of Variance (ANOVA), it was observed that exposure time and orientation play an important role. [ABSTRACT FROM AUTHOR]

Lately, accurate recognition of Process Control Chart Patterns (CCPs) has been considered one of the significant tactics for supervising manufacturing processes in order to achieve better control and to improve products quality. In addition to the single patterns, a huge work has been done to recognize concurrent ones that are usually due to the presence of two or more single patterns. Feature-based approaches are more efficient in pattern recognition especially in the case of concurrent patterns. Furthermore, a selection of an optimal set of features can significantly reduce the diagnostic search process. In this paper, a new approach based on a combination of Adaptive neuro" fuzzy inference system (ANFIS), Wavelet analysis (WA) and Principal components analysis (PCA) is used to recognize concurrent patterns ANFIS has proven to show high accuracy in pattern recognition. WA is used in this paper to improve the characteristics of the patterns to facilitate the recognition process by adding frequency features to the original pattern. Then, thirteen statistical features are extracted and PCA reduces their number to the most important three. ANFIS is used for training and testing the data based on the three extracted features as inputs and the patterns as target outputs. Extensive performance evaluation was carried out under normally and various non-normally distributed data. The non-normality of the inputs is based on two Gamma and Beta distributions. Results indicate that the proposed approach performs with high accuracy even in the case of non-normally distributed patterns. [ABSTRACT FROM AUTHOR]

Dynamic time warping (DTW) is a method of defining the similarity of waveforms. In contrast to cross-correlation, DTW allows signals to be stretched and compressed ("warped") in order to find the best match possible. Once a metric of similarity has been defined, a set of many signals can be partitioned into groups that are mutually similar. Separating signals into similar groups can be a useful first step in finding potential sources of noise. In this paper, the theory of dynamic time warping is reviewed with examples. DTW was used to partition recordings of transient percussion sounds into groups, with each group representing a source instrument. After partitioning, an "average" signal for each group was calculated, which was then used to compare the different sources of sound. [ABSTRACT FROM AUTHOR]

This article presents a new remote fault diagnosis and alarm architecture with the use of Cyber-physical systems and Cloud technology. Due to the large quantity of data collected from machines and their processing Cloud computing and Cyber-physical systems play a key role in such maintenance purposes. Currently, several remote diagnosis troubleshooters with different diagnostic tools and connection techniques have been proposed, but they rarely find interpretable diagnosis techniques. As such, the purpose of this paper is to propose a cyber-remote fault diagnosis system based on the combination between pattern recognition techniques and Cloud Technologies. The new troubleshooter is based on Cloud Computing to connect the system components achieving the cyber-physical system definition for Industry 4.0. Our system is a remote online device that measures vibration signals. The proposed troubleshooter is tested using the run-to-failure of aerospace rolling element-bearing data sets. The results reveal that the proposed methodology improves diagnostic accuracy compared with the most popular methodology used in other related systems. Finally, current and future Cyber-physical systems and Cloud technologies applications in the maintenance field are discussed. [ABSTRACT FROM AUTHOR]

Being part of a digital societies requires a "digital mindset". But what exactly is a digital mindset? There are many answers that have not really been systematized. A related area where systematization already started is Computational Thinking (CT): a branch of pedagogy that catalogued the capabilities necessary to live in a digitized and technology-driven world. CT is defined as a general problem-solving competence based on abstraction, algorithmic-thinking, problem-decomposition, pattern-recognition, and evaluation. Researchers promote CT as a 21st century skills important for everyone. Aim of our article is to reveal the relevance of CT as complement to IS-research in the creation of a digital mindset. We conducted a computational and qualitative literature review using quantitative bibliometric-citation analysis. From the analysis, we derive an integrative framework to identify how IS-research could profit from teaching, and applying CT, as a competence in the digital transformation of their discipline. [ABSTRACT FROM AUTHOR]

Artificial Intelligence (AI) is an umbrella term for systems that can act in cognitive processes in a human-like and human-enhancing manner, e.g., in learning, problem solving, and pattern recognition. According to models of technology adoption, several factors influence the actual implementation of a new system within an organization and in an individual's professional practice. These factors include e.g. job relevance, demonstrable results, individual experience with technology, and voluntariness to adopt the new system. This research studies employees' views and expectations of AI applicability and its impact to teachership within a Finnish higher education institution (HEI). Survey data was collected from different schools and units from all hierarchical layers of the HEI, a University of Applied Sciences. Views on AI were assessed in relation to the core tenets of a teacher's professional guidelines as expressed in the Comenius' Oath. This research contributes to the AI research by shedding light on how people within the HEI evaluate the impacts of AI into their future operating environment, pointing out also the potential obstacles for AI adoption in this specific context. [ABSTRACT FROM AUTHOR]

With the rise of Industry 4.0 and the implementation of "smart machines" new opportunities have opened to optimize the utilization of all data gathered from machines in an automotive manufacturing environment. The most prevalent model found in predictive maintenance is data-driven and based on statistical process control, pattern recognition or machine learning algorithms. A big obstacle faced by maintenance staff is understanding the data retrieved from the machines as the volume is often overwhelming and needs to be processed to be fully useful. Hence, this paper identifies the requirements for a successful implementation of predictive maintenance specifically for the use case of bolted connections using electronically controlled nutrunners in a modern, high-volume manufacturing environment. [ABSTRACT FROM AUTHOR]

In the medical area, the role of computer in pattern recognition of a disease is very much needed. It can be help for making treatment decision by first knowing the pattern of the disease. One of this model is the initial pattern recognition of burns that experienced by patients. Detecting the initial pattern of burn rates on the body will help the medical team immediately make decisions regarding patient burn level. To detect the initial pattern of burn rates, the appropriate method is to use the fisherface algorithm. This algorithm is used because of the ability to extract important information in imaging burn patterns on the body through the calculation of the average vector matrix and the covariance matrix in the pattern imaging database. In the process, the fisherface algorithm will generate an eigenface which is used for pattern recognition. Eigenface is the basis for calculating the burn patterns value on the body which represent the individual values for one or more pattern images of the burns on body. The using of fisherface and eigenface enable us to detect the degree of burns. The computational process will help in determining the degree of burns on body through facial recognition software. [ABSTRACT FROM AUTHOR]

Control chart pattern recognition (CCPR) is an essential tool for monitoring and diagnosing manufacturing process variability. It is used for recognizing manufacturing processes' abnormality. The specific type of patterns can be predicted with improved classification accuracy and less computational time when using appropriate features set in classifiers. Various features set extracted from process data streams have been proposed by researchers as input data representations for control chart pattern recognition (CCPR). This could confuse new researchers as to which features set need to be selected. Therefore, this paper aims to compare statistical features, shape features and mixed features as used in CCPR and identifies related open issues and research trends. This review concludes that mix features for input data representation are more promising to achieve a better recognition performance in terms of accuracy compared to the statistical and shape features. [ABSTRACT FROM AUTHOR]

Traditional Six Sigma curriculums including DMAIC, DFSS, DFR, Lean are not developed specifically and effectively for today's AI Data Science fields. This project will demonstrate an innovative Six Sigma training curriculum for Data Scientists. There are several objectives in this modern Six Sigma training curriculum: (1) adopt modern Text Mining and Data Mining techniques on Root Cause and Problem Solving Analyses such as DFMECA, C&E, QFD, SIPOC, VSM, (2) integrate various JMP statistical platforms holistically to analyze pattern recognition and discover the insights, (3) enhance predictive modeling capability through Neural, Partition, Principal Component algorithms, (4) utilize modern Quality and Process platforms such as Goal Plot, Model-Driven Multivariate SPC, Process History Process and Screening to enhance production quality control, (5) map these modern JMP Platforms into the Six Sigma DMAIC/DFSS/Lean framework for facilitating the Six Sigma Project execution, and (6) integrate database through Query Builder to monitor the production status in real-time. This curriculum is not just designed for a Data Scientist but is also powerful for those working in process, quality, reliability, supply chin, business, statistics, and marketing fields who want to become a reliable decision maker and true project leader. This Six Sigma DMAIC for Data Scientist Approach can be commonly applied to other Quality Engineering, Project Management and Business Excellence methodology such as DFSS, DFR, Lean, 8D, PMP. [ABSTRACT FROM AUTHOR]

Control charts are widely used in manufacturing to monitor whether a process is in control to ensure the consistency of the process. However, on emergence of any unexpected occurrence in a process, a control chart is unable to recognize the pattern of the dataset. Over the years, different methods have been developed to recognize unnatural control chart patterns. A methodology for recognizing control chart patterns is statistical correlation measure that measures correlation between a test dataset and a reference dataset. Support vector machine (SVM) has been the most popular model in recent years to recognize control chart patterns. However, computational effort in SVM model is high and training process of the model is complex and expensive. In this paper, these two methodologies have been applied to different datasets to recognize unnatural patterns. Their performance have been measured and compared. The comparison shows that SVM model performs substantially better than statistical correlation method in terms of accuracy in the output. Statistical correlation model is less expensive and associated with a simpler algorithm than the SVM model, but not as accurate as SVM model. On the other hand, the computational time required for SVM model is higher than the statistical correlation model. [ABSTRACT FROM AUTHOR]

Weight pruning is a powerful technique to realize model compression. We propose PCNN, a fine-grained regular 1D pruning method. A novel index format called Sparsity Pattern Mask (SPM) is presented to encode the sparsity in PCNN. Leveraging SPM with limited pruning patterns and non-zero sequences with equal length, PCNN can be efficiently employed in hardware. Evaluated on VGG-16 and ResNet-18, our PCNN achieves the compression rate up to 8.4 x with only 0.2% accuracy loss. We also implement a pattern-aware architecture in 55nm process, achieving up to 9.0 x speedup and 28.39 TOPS/W efficiency with only 3.1% on-chip memory overhead of indices. [ABSTRACT FROM AUTHOR]

Photonic integrated circuits suffer from large process variations. Effective and accurate characterization of the variation patterns is a critical task for enabling the development of novel techniques to alleviate the variation challenges. In this study, we propose a hierarchical approach that effectively decomposes the spatial variations of silicon microring-based optical transceivers into wafer-level, intra-die, and inter-die components. We then demonstrate that the characterized variation models can be used to generate trustworthy synthetic data for architecture- and system-level solutions for variation alleviation. We further demonstrate the utility of our variation characterization method for accurate yield prediction based on partial measurement data. [ABSTRACT FROM AUTHOR]

Nowadays more and more datacenters cooperate each others to achieve a common and more complex goal. New advanced functionalities are required to support experts during recovery and managing activities, like anomaly detection and fault pattern recognition. The proposed solution provides an active support to problem solving for datacenter management teams by providing automatically the root-cause of detected anomalies. The project has been developed in Bari using the datacenter ReCaS as testbed. Big Data solutions have been selected to properly handle the complexity and size of the data. Features like open source, big community, horizontal scalability and high availability have been considered and tools belonging to the Hadoop ecosystem have been selected. The collected information is sent to a combination of Apache Flume and Apache Kafka, used as transport layer, in turn delivering data to databases and processing components. Apache Spark has been selected as analysis component. Different kind of databases have been considered in order to satisfy multiple requirements: Hadoop Distributed File System, Neo4j, InfluxDB and Elasticsearch. Grafana and Kibana are used to show data in a dedicated dashboards. The Root-cause analysis engine has been implemented using custom machine learning algorithms. Finally, results are forwarded to experts by email or Slack, using Riemann. [ABSTRACT FROM AUTHOR]

One of the most important aspects of data processing at SHiP [1] experiments is tracks pattern recognition. The purpose of the SHiP Spectrometer Tracker (SST) is efficient reconstruction of charged particle tracks originating from decays of neutral New Physics objects. The reconstruction performance strongly depends on the tracker design and should be considered as an objective to define the best SST geometry parameters. In this study the SHiP Spectrom eter Tracker geometry optimization using Bayesian optimization with Gaussian processes in considered. The study have been done on MC data. The first results of the optimization are also considered. [ABSTRACT FROM AUTHOR]

Structural health monitoring (SHM) entails in situ sensing of a structure with the goal of detecting whether damage has occurred. The traditional SHM problem is a statistical pattern recognition problem, where the goal is to characterize structural damage by statistically correlating features extracted from sensor response data with data features of known damage. However, this typically requires a priori knowledge of how the structure will behave when it is damaged (e.g. a baseline state response). Model-based methods, however, do not require knowledge of the baseline structural response, but instead rely on a high-fidelity, physics-based model of the structural health monitoring system that is parameterized with features describing the damage to be characterized. The Markov chain Monte Carlo (MCMC) inverse solution entails recovering the damage parameter posterior probability distributions by repeatedly evaluating the forward model. However, utilizing physics-based models within the MCMC method has two main challenges: the damage posterior probability function changing if the structure changes and MCMC cannot be easily computationally parallelized (resulting in prohibitively long solution times or the use of less accurate surrogate models). This paper investigates the use of the Sequential Monte Carlo (SMC) method to addresses both of these issues. An SMC inverse solution is demonstrated on a six parameter crack characterization problem through a simulated ultrasonic guided wave inspection scenario consisting of an array of sensors. This paper also discusses the implementation of an optimized ultrasonic elastodynamic compute kernel for isotropic materials and its integration into the parallelized SMC parameter estimation method. The code is optimized to run efficiently on the Intel Xeon Phi Knights Landing manycore compute hardware. [ABSTRACT FROM AUTHOR]

Process Compensated Resonance Testing (PCRT) combines the collection of broadband resonance data with advanced pattern recognition to produce a fast, accurate, and automated non-destructive inspection for aerospace, automotive, and power generation components. To create a PCRT targeted defect inspection (Sorting Module) the resonance spectra of statistically significant populations of characterized acceptable and unacceptable parts are needed to train PCRT algorithms to recognize the frequency patterns that indicate defects in the midst of normal, acceptable material and geometry variations. In cases where a sufficient number of parts are not available, spectra from physical part populations can be supplemented with ’virtual’ spectra generated with PCRT forward models. Previous work investigated the creation of model-trained sorting modules for the detection of creep deformation and crystal orientation for coupon and turbine blade geometries made from single crystal Ni-based superalloy. In this work, model-trained Sorting Modules for crystal orientation and creep deformation were created and then validated with statistically significant populations of physical coupons with and without creep deformation, and with a range of crystal orientations. [ABSTRACT FROM AUTHOR]

E-learning platform quality is an important subject in the community of researchers and practitioners. Quality assessment approaches have different perspectives and differ in many respects, such as: objective, target groups and methods used. In recent years, the utility of advanced person-centered methods has become increasingly obvious. Person-centered approaches assume that samples come from heterogeneous populations and focus on identifying unobservable subpopulations that include individuals or similar cases. In this category are included both traditional techniques (e.g. cluster analysis) and advanced techniques (e.g. Latent Class Analysis, Latent Profile Analysis). In the present study, the e-learning platform quality is a general concept defined through three dimensions. The overall perception of students regarding the e-learning platform quality is based on the assessment of the following dimensions: e-learning system quality, information quality, and service quality. The aim of this paper is to extend previous research by adopting a person-oriented approach (e.g. LPA - latent profile analysis) and examine the heterogeneity of students' perceptions of e-learning platform quality. A sample of 385 students from five Romanian universities participated in the study. The data was analyzed with Mplus package version 7.4. The results of LPA revealed the presence of three profiles, reflecting different levels of e-learning platform quality perceptions: low, moderate, and high. The resulting profiles are distinct with respect to e-learning system quality, information quality, and service quality. A one-way multivariate analysis of variance (MANOVA) showed that the perceptions of e-learning platforms quality differed significantly across the three profiles. Finally, the article presented the practical implications of the results. The results contribute, on one hand, to identifying groups of students who shared similar perception patterns when dealing with e-learning platform quality. On the other hand, the results are useful in developing a strategy of improving the e-learning platforms by including the specific needs of each student categories identified. [ABSTRACT FROM AUTHOR]

Myoelectric pattern recognition has shown promise in the control of the prosthetic arm. But interfering noise and motion artifacts have hindered this kind of method to be used outside a controlled environment. This paper has designed an optimized circuit to process the Electromyography (EMG) signal. EMG signal is acquired from surface skin using sEMG (Surface Electromyography) electrode. EMG can be defined as an electrical potential produced due to contraction of the muscle. EMG signal requires to the undergo process of amplification and noise reduction before it can be converted to digital signal by the analog-digital converter (ADC) and processed to drive the motors or actuators of the prosthetic arm. The proposed denoising algorithm will improve the signal to noise ratio in real life uses. But the improved signal to noise ratio is expected to be insignificant as machine learning algorithm integrate noise as part of the signal. However, the variation of noise in real life uses is expected to occur where the proposed algorithm would potentially have a positive impact and further enhance the feasibility of using prosthetic for daily life [ABSTRACT FROM AUTHOR]

Wind speed prediction plays a very important role in the trading plan and reliable operation of power system. According to the formation mechanism and influencing factors of the wind, this paper presents a method to predict the wind speed by using adaptive network-based fuzzy inference system (ANFIS). ANFIS uses the hybrid learning algorithm to get the parameters of the front part and the conclusion of the training network, then inputs the selected wind speed samples into the trained adaptive fuzzy neural network, and finally forecasts the wind speed. [ABSTRACT FROM AUTHOR]

The goal of gesture recognition in Computer Science field has always been the minimization of the distance between the physical world and the digital world. The way humans interact among themselves could be implemented in communication with the digital world by interpreting gestures via mathematical algorithm. Numerous ways and algorithms have been proposed and implemented to achieve the goal of gesture recognition and its use in communicating with the digital world. Gestures can be tracked using accelerometers. a gesture controlled servo motor is controlled by using hand here one only needs to move hand to controlled the servo motor .A transmitting device is used in your hand which contains accelerometer. This will transmit command to servo motor. So that it can do the require task like moving forward, backward, and neutral. All this task will be performed by using hand gesture. [ABSTRACT FROM AUTHOR]

There are some relevant problems at prospecting-metallogenic analysis of perspective areas: 1. definition of expected formation (mineral) type of mineralization; 2. estimation of the scale of predicted ore mineralization; 3. outlining of the closest analogue from etalon deposits. During formalization this problems can be determined as the problems of pattern recognition, when objects under study described in heterogeneous property space. The conceptual approach to technology support is following: 1. prospecting estimations are carried out by standard models of ore objects of expected formation type of mineralization; 2. this models are developed using the system - hierarchic approach and logic-informative methods of data analysis; 3. the closest analogues are determined based on the calculation of the measure of similarity between objects. The logicinformative solution of recognition problem is based on determination of intervals – indicators for properties in each group of objects under study and consists of three phases: 1. definition of change limits of properties (intervals - indicators) describing groups of objects under study; 2. quantitative assessment of their self descriptiveness (dividing properties); 3. construction of rule for object reference to one of studying groups on the base of the obtained estimations. These methods allowed to solve reliably the outlined problems [ABSTRACT FROM AUTHOR]

The Mahalanobis-Taguchi System (MTS) refers to a newly-developed technique based on statistics that integrates a number of mathematical concepts to be applied for classification and diagnosis purposes within systems that are comprised of multiple dimensions. The MTS has been proven to be an exceptional technique that can be employed in numerous fields. In MTS, it is essential to choose the variables in order to enhance the accuracy in classifying via orthogonal array (OA) and Signal-to-Noise (S/N) ratios. However, the penalty for over-fitting or regularization is not included in the feature selection process for the MTS classifier. Hence, this paper investigated the combination between MTS and statistical pattern recognition approach applied to automotive crankshaft remanufacturing as an automated decision-making tool for classification purposes. The outcomes revealed that MTS displayed better classification performance for both training and test datasets, besides eliminating redundant and irrelevant parameters better than the conventional approach did. [ABSTRACT FROM AUTHOR]

One of the main concerns in the pattern recognition field is enhancing the predictive accuracy of the classification algorithms. While some efforts are focused in the development of new algorithms, another approach is centered in the search of the combination of data preparation techniques, a subset of features and, the selection of the adequate learning algorithm with its hyper parameters tuned. This approach is known as Full Model Selection (FMS) and because of the high search space, this paradigm is largely unexplored especially in large size datasets. In this work, the use of the meta-learning approach is suggested as a way to solve the FMS problem in huge datasets and retain the knowledge gained of each analysis performed which would be lost in another way. [ABSTRACT FROM AUTHOR]

The article describes methods of individual prediction. The methods of the Theory of Pattern Recognition were used for the models generation. The technique of the models generation and individual forecasting of electronic means for the space equipment has been modernized. The verification of derived models was carried out. The analysis of the models efficiency based on probabilistic characteristics was carried out. [ABSTRACT FROM AUTHOR]

The paper presents a new, developed by the author, concept of an artificial neural map, named COMSOM and inspired by classic SOM neural networks. The initial part of the article shows a general idea of COMSOM model and its structure, which includes two layers of artificial neurons: input layer (with typical task of distributing input values to data processing neurons) and basic layer containing usually a great number of processing elements, located inside a rectangular part of an XY plane, and able to move freely in this area during the learning stage. The Y axis of the rectangle represents an output variable (undergoing forecasting or evaluation) in a considered problem. A COMSOM training algorithm, proposed in next part of the paper, is an iterative machine learning procedure (based on special mathematical rules) and leads to both weight vector adaptation and moving neurons' positions along X and Y axes. In each learning step, only currently selected, relatively small number of neurons (belonging to a specifically determined neighbourhood like in a classic SOM networks) undergo learning. Finally, nodes equipped with weight vectors recognizing specific groups of objects (input feature vectors) are placed on proper X and Y positions, and can signalize proper value of output (forecasted or evaluated) variable as a value of Y coordinate. Such a model, after completing a training stage, realizes a projection of n-dimensional feature space (in which considered objects, or - actually - their feature vectors, are placed) into a 2-dimensional part of plane, with special projection of output variable into the Y axis. The model enables recognition and visualization of clusters of objects, with consideration of their output variable values. During the exploitation stage, when a new pattern is delivered on the input of trained COMSOM network, the "winning" neuron (that best recognizes object being processed) indicates not only the object membership to a given cluster (with visualization in the XY map), but also indicates the predicted output value on the Y axis. Promising results of application of a COMSOM model, in a problem of electricity load forecasting, has been presented in the last part of the paper. Final conclusion and directions of further research have been indicated. [ABSTRACT FROM AUTHOR]

This article focuses on development and design of class of the counterparty reliability recognition system which constitutes verification process of potential or currently existing counterparty of any scale and form of ownership. The Relevance of the development depends on organizations' necessity for understanding of due diligence indicators in contract award. The target audience of the program is small and medium enterprises of Russian Federation and individual entrepreneurs who can gain additional competitive advantages by using the developed service. The system is also intended to greater economic stability and commercial security of Russian enterprises. The information used in program preparation was gained from open related informational sources and expert materials in commercial security and jurisprudence. [ABSTRACT FROM AUTHOR]

The article presents an practical results of optimization of the fuzzy logic analysis method for pattern recognition. The theoretical background of the proposed theory is shown in the former article extending the original fuzzy logic analysis method. This article shows the implementation and experimental verification of the approach. [ABSTRACT FROM AUTHOR]

The article presents an optimization of the fuzzy logic analysis method for pattern recognition. The enhancements of the original method through the usage of additional two types of pattern components - leftwise diagonal and rightwise diagonal ones. The method is described in theoretical background and further articles show the implementation and experimental verification of the approach. [ABSTRACT FROM AUTHOR]

Opening the new firm or branch with desired execution is very relevant to facility location problem. Along the lines to locate the new ambulances and firehouses, the government desires to minimize average response time for emergencies from all residents of cities. So finding the best location is biggest challenge in day to day life. These type of problems were named as facility location problems. A lot of algorithms have been developed to handle these problems. In this paper, we review five algorithms that were applied to facility location problems. The significance of clustering in facility location problems is also presented. First we compare Fuzzy c-means clustering (FCM) algorithm with alternating heuristic (AH) algorithm, then with Particle Swarm Optimization (PSO) algorithms using different type of distance function. The data was clustered with the help of FCM and then we apply median model and min-max problem model on that data. After finding optimized locations using these algorithms we find the distance from optimized location point to the demanded point with different distance techniques and compare the results. At last, we design a general example to validate the feasibility of the five algorithms for facilities location optimization, and authenticate the advantages and drawbacks of them. [ABSTRACT FROM AUTHOR]

Head pose estimation has been widely used in the field of artificial intelligence, pattern recognition and intelligent human-computer interaction and so on. Good head pose estimation algorithm should deal with light, noise, identity, shelter and other factors robustly, but so far how to improve the accuracy and robustness of attitude estimation remains a major challenge in the field of computer vision. A method based on deep learning for pose estimation is presented. Deep learning with a strong learning ability, it can extract high-level image features of the input image by through a series of non-linear operation, then classifying the input image using the extracted feature. Such characteristics have greater differences in pose, while they are robust of light, identity, occlusion and other factors. The proposed head pose estimation is evaluated on the CAS-PEAL data set. Experimental results show that this method is effective to improve the accuracy of pose estimation. [ABSTRACT FROM AUTHOR]

Layout pattern classification, which groups similar layout clips into clusters, underlies a variety of design for manufacturability (DFM) applications such as hotspot library generation, hierarchical data storage, and yield optimization speedup. The key challenges of layout pattern classification are clip representation and clip clustering, while the mutually conflicting concerns are efficiency and solution quality (in terms of cluster count). In this paper, we present a fast and general layout pattern classification algorithm. Our simple but general clip representation captures both topology and density; we can handle not only rigid area match or edge displacement constraints but also variant edge tolerances and don't care regions. On the other hand, for achieving a small cluster count, our clip clustering is guided by the natural grouping structure of layout clips. Our experiments are conducted on 2016 CAD contest at ICCAD benchmark suite; our results show that our algorithm outperforms the reference solution and all contest winning teams, delivering the smallest cluster count, fastest runtime, and 100% validity. In addition to the good solution quality, the interplay between adopted simple and easily manipulated data structures and our algorithm makes it fast and viable to be incorporated into practical DFM flows. [ABSTRACT FROM AUTHOR]

Reorderable matrices are the basis of heatmaps and other tabular displays in which row and column permutations are possible. Matrix reordering algorithms try to provide a permuted version of these matrices, such that hidden patterns become visible for users. However, most of them fail on revealing Circumplex pattern, due to high execution time or to low quality of output matrix. In order to overcome this difficulty, we present a matrix reordering method (Circumplex Sort), whose goal is to permute a matrix in order to reveal a Circumplex pattern. In a first phase, our method reorders input matrix rows according to the index-weighted means of the values of each column. After that, the method creates and concatenates two stacks of rows based on a same initial row; for each row obtained according to the row order of the first phase, the method pushes this row at the stack whose top is most similar to it. Circumplex Sort runs the same two phases for columns. Our experiments with synthetic matrices revealed that our approach produces high-quality results (in terms of evaluation functions and runtime) when this pattern is hidden in data matrix. Besides, we present a real world dataset that exemplifies the use of Circumplex Sort for revealing Circumplex pattern. [ABSTRACT FROM AUTHOR]

Automatic recognition of situations with moving objects in videos is crucial in various application fields. Trajectory-based methods for recognition of situations with moving objects in videos include finite state machines, hidden Markov models, dynamic time warping, etc. Methods based on dynamic programming like hidden Markov models and dynamic time warping are robust under noisy conditions but require preparation of training datasets that can be difficult in case of rare situations. Methods based on finite state machines make it possible to eliminate training stage at all but lack robustness under noisy conditions. This paper introduces a novel method of situation recognition that combines advantages of finite state machines and dynamic programming. The proposed method was benchmarked on recognition of situations with moving objects in road traffic scenarios. Experimental results confirmed efficiency of the proposed method on real videos. [ABSTRACT FROM AUTHOR]

We report on the performance of a specialized processor capable of reconstructing charged particle tracks in a realistic LHC silicon tracker detector, at the same speed of the readout and with sub-microsecond latency. The processor is based on an innovative pattern-recognition algorithm, called "artificial retina algorithm", inspired from the vision system of mammals. A prototype of the processor has been designed, simulated, and implemented on Tel62 boards equipped with high-bandwidth Altera Stratix III FPGA devices. The prototype is the first step towards a real-time track reconstruction device aimed at processing complex events of high-luminosity LHC experiments at 40 MHz crossing rate. [ABSTRACT FROM AUTHOR]

PANDA is one of the four experiments that will run at the new facility FAIR that is being built in Darmstadt, Germany. It is a fixed target experiment: a beam of antiprotons collides on a jet proton target (the maximum center of mass energy is 5.46 GeV). The interaction rate at the startup will be 2MHz with the goal of reaching 20MHz at full luminosity. The beam of antiprotons will be essentially continuous. PANDA will have NO hardware trigger but only a software trigger, to allow for maximum flexibility in the physics program. All those characteristics are severe challenges for the reconstruction code that 1) must be fast, since it has to be validated up to 20MHz interaction rate; 2) must be able to reject fake tracks caused by the remnant hits, belonging to previous or later events in some slow detectors, for example the straw tubes in the central region. The Pattern Recognition (PR) of PANDA will have to run both online to achieve a first fast selection, and offline, at lower rate, for a more refined selection. In PANDA the PR code is continuously evolving; this contribution shows the present status. I will give an overview of three examples of PR following different strategies and/or implemented on different hardware (FPGA, GPUs, CPUs) and, when available, I will report the performances. [ABSTRACT FROM AUTHOR]

Linear discriminant analysis (LDA) is a generalization of Fisher's linear discriminant, a method used in statistics, pattern recognition and machine learning to find a linear combination of features that characterizes or separates two or more classes of objects or events. It is one of the methods offered in multivariable analysis can be helping in land classification. This research resulted functions and values discriminant that could be used for classification of sub-district land in Banda Aceh by using Fisher discriminant analysis. The functions and values discriminant that is obtained had been implemented on vacant land data in sub-district of Baiturrahman. Results of this research are sub-district of Baiturrahman having 340 vacant lands, by which 282 vacant lands are used as protected areas and 58 others are used as cultural areas. Vacant land area in sub-district of Baiturrahman is 40.54 Ha, where land area that can be used as protected area is 32.216 Ha and land area that can be used as cultural areas is 8.324 Ha. Furthermore, discriminant function that is obtained has the accuracy of classifying land of 62.82%. This research is expected to be a recommendation to classify vacant land in Banda Aceh city, vacant land use for sub-district of Baiturrahman can be a consideration to government, so that further research can be done to classify vacant land directly into 11 land use in Banda Aceh city. [ABSTRACT FROM AUTHOR]

Batik is one of the world's cultural heritages with various decorative patterns. Each batik has its unique pattern characteristic such as color intensity, ornament visualisation and ornament size. Batik with a large-sized ornament and a high-intensity color on its background will result in a clear edge and easy-to-recognise ornaments. While batik with a small-sized ornament and a low-intensity color of the background will produce slight edge ornaments, thus it becomes difficult to recognise. In order to recognise any patterns of batik, some features that can represent each pattern are becomes necessary. Batik pattern features can be obtained from various methods of feature extraction. This paper will address various methods of feature extraction of batik for recognising batik patterns. These methods of feature extraction are categorised into several groups based on visual features produced by batik including feature extraction of color, texture formed in batik, and shape of batik ornament. The grouping of extraction methods in batik aims to describe the methods for extracting features from batik pattern to improve the recognition level of batik pattern. [ABSTRACT FROM AUTHOR]

The statistical data of aortic valve is very important input information for medical diagnostics and anatomical pathology research. The approach for this data extraction is photo survey of aortic valve and next image processing. In this paper the authors have implemented different methods of pattern recognition for extracting the statistical data from photographic images. The modification of "k-Nearest Neighbors" method was proposed. The experimental results show the efficiency of the selected methods and proposed modification. [ABSTRACT FROM AUTHOR]

In this paper pattern recognition problem with rejecting option is discussed. The problem is aimed at classification patterns from given classes (native patterns) and rejecting ones not belonging to these classes (foreign patterns). In practice the characteristics of the native patters are given, while no information about foreign ones is known. A rejecting tool is aimed at enclosing native patterns in compact geometrical figures and excluding foreign ones from them. [ABSTRACT FROM AUTHOR]

Patterns are mentioned usually in the extraction context. Little stress is posed in their representation and management. This paper is focused on representation of the patterns, manipulation with patterns and query patterns. Crucial issue can be seen in systematic approach to pattern management and specific pattern query language which takes into consideration semantics of patterns.. [ABSTRACT FROM AUTHOR]

The problem of pattern recognition has been solved with numerous methods in the Artificial Intelligence field. We present an unconventional method based on Lingustic Fuzzy Logic Forecaster which is primarily used for the task of time series analysis and prediction through logical deduction wtih linguistic variables. This method should be used not only to the time series prediction itself, but also for recognition of patterns in a signal with seasonal component. [ABSTRACT FROM AUTHOR]

Cluster analysis or clustering is a task of grouping a set of objects in such a way that objects in the same group (called a cluster) are more similar (in some sense or another) to each other than to those in other groups (clusters). It is the main task of exploratory data mining and a common technique for statistical data analysis used in many fields, including machine learning, pattern recognition, image analysis, information retrieval, and bioinformatics. The topic of this paper is one of the modern methods of clustering namely SOM (Self Organising Map). The paper describes the theory needed to understand the principle of clustering and descriptions of algorithm used with clustering in our experiments. [ABSTRACT FROM AUTHOR]

Hand gestures can easily express ideas to others and these hand gestures can be applied to various fields such as art, game, and smart home. In this study, we propose a method to improve recognition rate for various kinds of 3-dimensional hand gestures. We extract the features from the gesture data distribution by Mahalanobis distance and construct the feature nodes, and feature nodes are added to the input layer of a neural network. Experimental results show that the proposed method has better performance than the conventional artificial neural network. [ABSTRACT FROM AUTHOR]

Designing serious games for learning proves to be a very difficult task, as it requires knowledge not only about the content, but also about game design and pedagogy. There have been efforts to create collections of design patterns that help to create initial designs or improve existing ones. However those collections neglect the role of social interactions within serious games for learning, even though this could lead to improvements in learning. This is both due to technical and design reasons. In this paper we want to give an overview over social interaction patterns observed in existing serious games. They are meant as a starting point to introduce social interactions into serious games. One pattern is used as an example to demonstrate the patterns approach. [ABSTRACT FROM AUTHOR]