Set Of Training Patterns Research Articles

Multi-Spectral Satellite Image Classification Using an Evolving Neural Network Approach

This paper investigates a new evolving neural network classifier based on real coded genetic algorithm for automatic multi-spectral satellite image classification (land cover mapping problem). The evolving neural network classifier is designed using hybrid genetic operators with classification accuracy as a measure of performance. The evolving neural network methodology is implemented in Pentium clusters. The proposed methodology searches for the best neural network architecture and its connection weights for a given set of training patterns. The performance of the proposed evolving neural network based classifier is evaluated for Level-II classifier model using the Landsat 7 Thematic Mapper high resolution imagery. After evolving the neural network at pixel level, the system performance is tested with sites not seen during training. Results are compared with maximum likelihood classifier, gradient based fully connected multilayer perceptron and growing and pruning radial basis function classifier. The proposed classifier is more accurate, robust with respect to the noise in the input spectrum and also overcomes the common limitations of the standard neural based classifier models.

A New Non-stationary High-order Spatial Sequential Simulation Method

A new non-stationary, high-order sequential simulation method is presented herein, aiming to accommodate complex curvilinear patterns when modelling non-Gaussian, spatially distributed and variant attributes of natural phenomena. The proposed approach employs spatial templates, training images and a set of sample data. At each step of a multi-grid approach, a template consisting of several data points and a simulation node located in the center of the grid is selected. To account for the non-stationarity exhibited in the samples, the data events decided by the conditioning data are utilized to calibrate the importance of the related replicates. Sliding the template over the training image generates a set of training patterns, and for each pattern a weight is calculated. The weight value of each training pattern is determined by a similarity measure defined herein, which is calculated between the data event of the training pattern and that of the simulation pattern. This results in a non-stationary spatial distribution of the weight values for the training patterns. The proposed new similarity measure is constructed from the high-order statistics of data events from the available data set, when compared to their corresponding training patterns. In addition, this new high-order statistics measure allows for the effective detection of similar patterns in different orientations, as these high-order statistics conform to the commutativity property. The proposed method is robust against the addition of more training images due to its non-stationary aspect; it only uses replicates from the pattern database with the most similar local high-order statistics to simulate each node. Examples demonstrate the key aspects of the method.

Generativ rəqib şəbəkələrin analizi

Recently, a lot of research has been done on the use of generative models in the field of computer vision and image classification. At the same time, effective work has been done with the help of an environment called generative adversarial networks, such as video generation, music generation, image synthesis, text-to-image conversion. Generative adversarial networks are artificial intelligence algorithms designed to solve the problems of generative models. The purpose of the generative model is to study the set of training patterns and their probable distribution. The article discusses generative adversarial networks, their types, problems, and advantages, as well as classification and regression, segmentation of medical images, music generation, best description capabilities, text image conversion, video generation, etc. general information is given. In addition, comparisons were made between the generative adversarial network algorithms analyzed on some criteria.

Information Fusion for Biological Prediction

Information fusion has become a powerful tool for challenging applications such as biological prediction problems. In this paper, we apply a new information-theoretical fusion technique to HIV-1 protease cleavage site prediction, which is a problem that has been in the focus of much interest and investigation of the machine learning community recently. It poses a difficult classification task due to its high dimensional feature space and a relatively small set of available training patterns. We also apply a new set of biophysical features to this problem and present experiments with neural networks, support vector machines, and decision trees. Application of our feature set results in high recognition rates and concise decision trees, producing manageable rule sets that can guide future experiments. In particular, we found a combination of neural networks and support vector machines to be beneficial for this problem.

A neuro-genetic Hopfield associative memory (HAM) with improved noise immunity

Artificial Neural Network (ANN) is a form of computation inspired by the arrangement and purpose of the brain. Pattern Recognition is one of the demanding areas under Artificial Neural Network and pattern association is one of the major tasks through which pattern recognition is realized. Nowadays researchers are taking interest in the combination of evolutionary algorithms and ANN. In the present work, an attempt is being made to analyze the performance of Hopfield Associative Memory with Monte Carlo adaptation rule and evolutionary searching (Genetic Algorithm) for pattern storage and recall. The training pattern set has been taken a set of five Greek alphabet letters as different objects. The aim is to store Patterns with MC adaptation rule and get the best weight matrices for efficient recalling of any given set patterns. The neural networks architectures have been taken up for storing the given set of inputs, for which conventional Hebbian and Monte Carlo adaptation rule is considered. During the mutation process, various mutation probabilities are supposed to be evaluated. The detailed simulated results thus obtained will be presented with the help of tables and graph.

Combining Neural Networks for Skin Detection

Two types of combining strategies were evaluated namely combining skin features and combining skin classifiers. Several combining rules were applied where the outputs of the skin classifiers are combined using binary operators such as the AND and the OR operators, Voting, Sum of Weights and a new neural network. Three chrominance components from the YCbCr colour space that gave the highest correct detection on their single feature MLP were selected as the combining parameters. A major issue in designing a MLP neural network is to determine the optimal number of hidden units given a set of training patterns. Therefore, a coarse to fine search method to find the number of neurons in the hidden layer is proposed. The strategy of combining Cb/Cr and Cr features improved the correct detection by 3.01% compared to the best single feature MLP given by Cb-Cr. The strategy of combining the outputs of three skin classifiers using the Sum of Weights rule further improved the correct detection by 4.38% compared to the best single feature MLP.

Neural Network-Based Supplementary Frequency Controller for a DFIG Wind Farm

An artificial neural network (ANN)-based supplementary frequency controller is designed for a doubly fed induction generator (DFIG) wind farm in a local power system. Since the optimal controller gain that gives highest the frequency nadir or lowest peak frequency is a complicated nonlinear function of load disturbance and system variables, it is not easy to use analytical methods to derive the optimal gain. The optimal gain can be reached through an exhaustive search method. However, the exhaustive search method is not suitable for online applications, since it takes a long time to perform a great number of simulations. In this work, an ANN that uses load disturbance, wind penetration, and wind speed as the inputs and the desired controller gain as the output is proposed. Once trained by a proper set of training patterns, the ANN can be employed to yield the desired gain in a very efficient manner, even when the operating condition is not included in the training set. Therefore, the proposed ANN-based controller can be used for real-time frequency control. Results from MATLAB/SIMULINK simulations performed on a local power system in Taiwan reveal that the proposed ANN can yield a better frequency response than the fixed-gain controller.

Robust Distance Measures for kNN Classification of Cancer Data.

The k-Nearest Neighbor (kNN) classifier represents a simple and very general approach to classification. Still, the performance of kNN classifiers can often compete with more complex machine-learning algorithms. The core of kNN depends on a “guilt by association” principle where classification is performed by measuring the similarity between a query and a set of training patterns, often computed as distances. The relative performance of kNN classifiers is closely linked to the choice of distance or similarity measure, and it is therefore relevant to investigate the effect of using different distance measures when comparing biomedical data. In this study on classification of cancer data sets, we have used both common and novel distance measures, including the novel distance measures Sobolev and Fisher, and we have evaluated the performance of kNN with these distances on 4 cancer data sets of different type. We find that the performance when using the novel distance measures is comparable to the performance with more well-established measures, in particular for the Sobolev distance. We define a robust ranking of all the distance measures according to overall performance. Several distance measures show robust performance in kNN over several data sets, in particular the Hassanat, Sobolev, and Manhattan measures. Some of the other measures show good performance on selected data sets but seem to be more sensitive to the nature of the classification data. It is therefore important to benchmark distance measures on similar data prior to classification to identify the most suitable measure in each case.

Image-based textile decoding

A textile fabric consists of countless parallel vertical yarns (warps) and horizontal yarns (wefts). While common looms can weave repetitive patterns, Jacquard looms can weave the patterns without repetition restrictions. A pattern in which the warps and wefts cross on a grid is defined in a binary matrix. The binary matrix can define which warp and weft is on top at each grid point of the Jacquard fabric. The process can be regarded as encoding from pattern to textile. In this work, we propose a decoding method that generates a binary pattern from a textile fabric that has been already woven. We could not use a deep neural network to learn the process based solely on the training set of patterns and observed fabric images. The crossing points in the observed image were not completely located on the grid points, so it was difficult to take a direct correspondence between the fabric images and the pattern represented by the matrix in the framework of deep learning. Therefore, we propose a method that can apply the framework of deep learning viau the intermediate representation of patterns and images. We show how to convert a pattern into an intermediate representation and how to reconvert the output into a pattern and confirm its effectiveness. In this experiment, we confirmed that 93% of correct pattern was obtained by decoding the pattern from the actual fabric images and weaving them again.

Automated Burned Scar Mapping Using Sentinel-2 Imagery

The Sentinel-2 satellites are providing an unparalleled wealth of high-resolution remotely sensed information with a short revisit cycle, which is ideal for mapping burned areas both accurately and timely. However, the high detail and volume of information provided actually encumbers the automation of the mapping process, at least for the level of automation required to map systematically wildfires on a national level. This paper proposes a fully automated methodology for mapping burn scars using Sentinel-2 data. Information extracted from a pair of Sentinel-2 images, one pre-fire and one post-fire, is jointly used to automatically label a set of training patterns via two empirical rules. An initial pixel-based classification is derived using this training set by means of a Support Vector Machine (SVM) classifier. The latter is subsequently smoothed following a multiple spectral-spatial classification (MSSC) approach, which increases the mapping accuracy and thematic consistency of the final burned area delineation. The proposed methodology was tested on six recent wildfire events in Greece, selected to cover representative cases of the Greek ecosystems and to present challenges in burned area mapping. The lowest classification accuracy achieved was 92%, whereas Matthews correlation coefficient (MCC) was greater or equal to 0.85.

Adaptive and Neural Network Based Control of Pitch of Unmanned Aerial Vehicles

The objective is to develop a new control strategy for quadrotor types of Unmanned Aerial Systems (UAVs) that assist in closing the existing research gaps between the undesired uncertainties and current control systems. This research investigates a system modeling and the effectiveness of error elimination through adaptation in the change of the plant/system’s output. The UAV is controlled via Wi-Fi using MATLAB. Proportional-Integral-Derivative (PID) controller, the two degrees of freedom Proportional-Integral-Derivative (2DOF-PID) controller and the model reference Neural Network (NN) controller are used for the aircraft pitch control. External disturbances and modeling errors can lead to the instability and unpredictable behaviors of a system. Even though the disturbances cannot be avoided the effects of such disturbances can be reduced significantly by neural network-based adaptive controllers for high-performance tracking in the presence of the disturbances. The robust control systems with the adaptive control will improve navigation performance and operation for military, reconnaissance and surveillance applications. The application of a robust controller determines the performance by observing the system’s behavior to identify and reduce the unpredictable effects of disturbances and achieve a better behavior of the whole system. By observing the system, the neural network is suitably used to learn from a set of training patterns. System identification uses a neural network to capture the behavior of system dynamics, which assists the neural network to train itself to act as a controller. Moreover, the system’s performance of the controllers was tested using simulations to demonstrate the effectiveness in improving the speed and stability for the dynamic system. The presented work will produce a flexible, robust and effective control system model that provides additional stability and reduce the effects of disturbances.

A database-driven neural computing framework for classification of vertical jump patterns of healthy female netballers using 3D kinematics–EMG features

Classification of athletes’ performance in vertical jump (VJ) tests is a recommended practice at various stages of athlete fitness development and performance enhancement. The practice is, however, currently conducted subjectively and mainly based on measures of the vertical jump height (VJH) attained in standardized VJ tests. The current study presents an intelligent integrated classification framework (IICF) for classification of athlete performance in single-leg (SL) and double-leg (DL) standing VJ tests based on lower extremity (LE) biomechanical data. Biomechanical data consisted of three-dimensional (3D) kinematics and electromyography (EMG) features generated from ankle–knee joints and eight LE muscles of 13 healthy female national netball players (subjects) obtained during six trials of SL-left leg (SLLL), SL-right leg (SLRL), and DL VJ tests. Each participating subject had prior VJ classification by the trainer as either excellent or very good in each of the three VJ tests. IICF introduced in this work utilizes an integration of the scalable and interoperable relational database management system (MySQLDB) and artificial neural networks (ANNs). Integrated pattern sets containing the extracted features (EF) obtained from first three VJ trials data were randomly partitioned into design and test data sets and ANN implemented using fully connected multilayer perceptron feedforward neural networks (MLP-FFNN) with three different training algorithms. Subjects’ prior VJ classifications were used as MLP-FFNN target outputs. A second classifier is trained using support vector machine (SVM) using three different kernel mapping functions and performances between MPL-FFNN and SVM classifiers compared on both training and independent test pattern sets. The test pattern sets comprise of EF generated from the latter three VJ trials data stored in MySQLDB. The average classification accuracy (F-measure) achieved by the optimally trained MLP-FFNN classifier on independent test pattern sets across all the three VJ activities was 93.33% (86.67–96.77%), whereas SVM classifier’s was 82.5% (73.33–87.5%). Through a custom-made web-based interface having backend integration of MySQLDB and the optimally trained MLP-FFNN, classification of individual subjects’ pattern sets from either VJ activities is also enabled. Individual subjects’ classification results can then be compared with prior classifications by the trainer, and differences were highlighted. IICF introduced in this work has demonstrated its feasibility as an objective complementary assessment tool for trainers during conducting of SL and DL vertical jump tests for netball players and athletes in general. Descriptive statistics for the entire experimental data sets’ features used in this study are also presented.

Performance evaluation of Hindi speech recognition system using optimized filterbanks

An Automatic Speech Recognition (ASR) system implementation uses a conventional pattern recognition technique that stores a set of training patterns in classes and compares the test patterns with training patterns to place them in the best matched pattern class. Most state-of-the-art ASR systems use Mel Frequency Cepstral Coefficient (MFCC) and Perceptual Linear Prediction (PLP) to extract features in training phase of the ASR system. However, sensitivity of MFCC & PLP to background noise has resulted in use of noise robust features Gammatone Frequency Cepstral Coefficient (GFCC) and Basilar-membrane Frequency-band Cepstral Coefficient (BFCC). But many issues associated with these feature extraction methods, like accepted bandwidth and standard number of filters are unresolved till date. This paper proposes a novel approach to use Differential Evolution (DE) algorithm to optimize the number and spacing of filters used in MFCC, GFCC and BFCC techniques. It also evaluates the performance of the said feature extraction methods with and without DE optimization in clean as well as in noisy environments. The results conclude that BFCC based ASR systems performs 0.4% to 1.0% better than GFCC and 7% to 10% better than MFCC in different conditions.

Facial expression recognition using a combination of multiple facial features and support vector machine

This paper presents a novel facial expression recognition (FER) technique based on support vector machine (SVM) for the FER. Here it is called the FERS technique. First, the FERS technique develops a face detection method that combines the Haar-like features method with the self-quotient image (SQI) filter. As a result, the FERS technique possesses better detection rate because the face detection method gets more accurate in locating face regions of an image. The main reason is that the SQI filter can overcome the insufficient light and shade light. Subsequently, three schemes, the angular radial transform (ART), the discrete cosine transform (DCT) and the Gabor filter (GF), are simultaneously employed in the design of the feature extraction for facial expression in the FERS technique. More specifically, they are employed in constructing a set of training patterns for the training of an SVM. The FERS technique then exploits the trained SVM to recognize the facial expression for a query face image. Finally, experimental results show that the recognition performance of the FERS technique can be better than that of other existing methods under consideration in the paper.

A Time Series Forecasting Method

This paper proposes a novel time series forecasting method based on a weighted self-constructing clustering technique. The weighted self-constructing clustering processes all the data patterns incrementally. If a data pattern is not similar enough to an existing cluster, it forms a new cluster of its own. However, if a data pattern is similar enough to an existing cluster, it is removed from the cluster it currently belongs to and added to the most similar cluster. During the clustering process, weights are learned for each cluster. Given a series of time-stamped data up to time t, we divide it into a set of training patterns. By using the weighted self-constructing clustering, the training patterns are grouped into a set of clusters. To estimate the value at time t + 1, we find the k nearest neighbors of the input pattern and use these k neighbors to decide the estimation. Experimental results are shown to demonstrate the effectiveness of the proposed approach.

Simultaneous face and facial expression recognition using a MLP neural network based on constructive training algorithm

This paper proposes a system for simultaneous recognition of faces and facial expressions using a constructive training algorithm. A MLP classifier has been developed to recognise at the same time face and facial expression. Feature extraction step is based on perceived facial images. The algorithm is formed by a reduced number of hidden neurons and a set of training patterns. A new hidden neuron is added when mean square error (MSE) of training data (TD) is not reduced to a predefined value. Input patterns are trained incrementally until all patterns of TD are learned. The proposed algorithm seeks to find synthesis parameters as patterns number corresponding for the subsets of each class to be presented initially in training step, initial number of hidden neurons, iterations number as well as the MSE value. By comparing with a fixed MLP and feature extraction techniques, the effectiveness of the proposed approach has been proven.

Simultaneous face and facial expression recognition using a MLP neural network based on constructive training algorithm

This paper proposes a system for simultaneous recognition of faces and facial expressions using a constructive training algorithm. A MLP classifier has been developed to recognise at the same time face and facial expression. Feature extraction step is based on perceived facial images. The algorithm is formed by a reduced number of hidden neurons and a set of training patterns. A new hidden neuron is added when mean square error (MSE) of training data (TD) is not reduced to a predefined value. Input patterns are trained incrementally until all patterns of TD are learned. The proposed algorithm seeks to find synthesis parameters as patterns number corresponding for the subsets of each class to be presented initially in training step, initial number of hidden neurons, iterations number as well as the MSE value. By comparing with a fixed MLP and feature extraction techniques, the effectiveness of the proposed approach has been proven.

Crack identification method for tapered cantilever pipe-type beam using natural frequencies

A method to detect location and size of a crack in tapered cantilever pipe-type beam using changes in natural frequencies is presented. The boundary conditions of cantilever beam are applied to a general solution for vibrating tapered beam. Then, an equivalent bending stiffness for cracked beam is used to obtain the natural frequencies. Numerical simulations are carried out to construct a set of training patterns of a neural network, and committee of neural networks is employed to identify the crack. Crack identifications are carried out for the 16 damage cases numerically and for the 3 damage cases experimentally. The identified crack locations and sizes agree reasonably well with the exact values.

A Nom historical document recognition system for digital archiving

A Nom historical document recognition system is being developed for digital archiving that uses image binarization, character segmentation, and character recognition. It incorporates two versions of off-line character recognition: one for automatic recognition of scanned and segmented character patterns (7660 categories) and the other for user handwritten input (32,695 categories). This separation is used since including less frequently appearing categories in automatic recognition increases the misrecognition rate without reliable statistics on the Nom language. Moreover, a user must be able to check the results and identify the correct categories from an extended set of categories, and a user can input characters by hand. Both versions use the same recognition method, but they are trained using different sets of training patterns. Recursive X---Y cut and Voronoi diagrams are used for segmentation; k---d tree and generalized learning vector quantization are used for coarse classification; and the modified quadratic discriminant function is used for fine classification. The system provides an interface through which a user can check the results, change binarization methods, rectify segmentation, and input correct character categories by hand. Evaluation done using a limited number of Nom historical documents after providing ground truths for them showed that the two stages of recognition along with user checking and correction improved the recognition results significantly.

Selecting patterns and features for between- and within- crop-row weed mapping using UAV-imagery

This paper approaches the problem of weed mapping for precision agriculture, using imagery provided by Unmanned Aerial Vehicles (UAVs) from sunflower and maize crops. Precision agriculture referred to weed control is mainly based on the design of early post-emergence site-specific control treatments according to weed coverage, where one of the most important challenges is the spectral similarity of crop and weed pixels in early growth stages. Our work tackles this problem in the context of object-based image analysis (OBIA) by means of supervised machine learning methods combined with pattern and feature selection techniques, devising a strategy for alleviating the user intervention in the system while not compromising the accuracy. This work firstly proposes a method for choosing a set of training patterns via clustering techniques so as to consider a representative set of the whole field data spectrum for the classification method. Furthermore, a feature selection method is used to obtain the best discriminating features from a set of several statistics and measures of different nature. Results from this research show that the proposed method for pattern selection is suitable and leads to the construction of robust sets of data. The exploitation of different statistical, spatial and texture metrics represents a new avenue with huge potential for between and within crop-row weed mapping via UAV-imagery and shows good synergy when complemented with OBIA. Finally, there are some measures (specially those linked to vegetation indexes) that are of great influence for weed mapping in both sunflower and maize crops.