In deep learning, understanding the decision-making processes of complex models is essential for advancing interpretability and trust in artificial intelligence systems. We introduce Causal Relational Attribution Graph (C-RAG), designed to deliver comprehensive, multi-perspective explanations of convolutional neural networks (CNNs) via a graph representation. C-RAG integrates gradient-based local attribution with global feature importance by constructing a graph-based representation that captures hierarchical feature inter-dependencies. In this framework, feature clusters are represented as graph nodes, and their interactions are quantified through combined localized and global attribution metrics, ensuring interpretable insights into model behavior. We evaluate C-RAG across diverse benchmark datasets (ImageNet, CIFAR-10, MNIST) and CNN architectures (ResNet18, VGG19, DenseNet201, LeNet), demonstrating significant advancements over state-of-the-art explainability methods in faithfulness, robustness, and computational efficiency. The proposed approach facilitates accurate spatial feature localization, robust dependency mapping, and efficient explanation generation, making it a valuable tool for critical applications such as medical imaging and autonomous systems. We provide a novel graph-based explainability framework, which bridges the gap between local and global interpretability, C-RAG addresses key limitations in existing methods, establishing a robust foundation for explainable AI in computer vision.

Abstract Deep Neural Networks (DNNs) have achieved state-of-the-art performance across various domains, yet their widespread adoption remains constrained by substantial computational and memory demands. While model pruning has emerged as a compelling strategy to address these challenges, existing methods often suffer from critical shortcomings: (1) non-selective neuron pruning which overlooks neuron activation dynamics, leading to the removal of critical neurons; (2) an inability to account for task-specific neuron importance, which causes accuracy degradation; and (3) failure to mitigate redundancy in neuron activations, resulting in suboptimal compression. In this work, we propose a novel Neuron Efficiency Metric (NEM), which integrates three key components—Neuron Activation Rate (NAR), Class-specific Activation Strength (CAS), and Neuron Overlap Index (NOI)—to address these limitations and guide a more effective pruning process. By iteratively evaluating the relevance of each neuron along these axes, NEM ensures selective and structured pruning that minimizes the retention of redundant or irrelevant neurons while preserving task-critical activations. The proposed method is tested on modern architectures using benchmark datasets such as MNIST and CIFAR-10, demonstrating a significant reduction in computational complexity while maintaining or even improving model performance. The results reveal that NEM achieves a higher degree of compression with minimal accuracy loss compared to conventional techniques.

Abstract Convolutional neural networks have become prominent machine learning models, particularly in the realm of computer vision, due to their ability to predict and extract robust features from raw image data. CNNs, similar to other neural network models, undergo training via backpropagation, an iterative technique. However, the backpropagation algorithm has notable challenges, including slow convergence, susceptibility to local minima, and hypersensitivity to learning rates. These challenges not only impact the model’s accuracy but also make the training process computationally intensive. To address these limitations, We introduce a novel approach that trains the CNN classifier using a non-iterative learning method. The proposed approach involves automatic extraction of pertinent features from the raw-data, followed by the application of Gram–Schmidt process to decompose the feature matrix and determine classifier’s weights. The proposed method has shown enhanced predictive accuracy over state-of-the-art models when evaluated on two benchmark datasets, MNIST and CIFAR-10. The extensive experimentation using most cited pre-trained experiments validate the effectiveness of our proposed method.

Deep Neural Networks (DNNs) have undeniably achieved groundbreaking success across diverse applications. Nevertheless, their complex architectures inherently lead to substantial computational demands and memory prerequisites. To surmount these challenges, this research paper introduces a pioneering approach designed to amplify DNN efficiency via a unique iterative pruning technique Neuron Efficiency Index (NEI), that considers activation frequency of each neuron, class sensitivity and redundancy in the dense layer neurons. The central objective of this method is to curtail the computational burden of the model, all the while ensuring that performance remains intact and enhanced. The proposed technique is used to prune state-of-the-art architectures and comprehensive comparison is presented on benchmark dataset MNIST and CIFAR-10. The evaluation presents that proposed NEI improves the model accuracy while reducing the computations and complexity of the architecture. The work contributes to the field of neural network optimization.

The modern deep learning-based architectures have performed well for pixel-wise segmentation tasks. The consideration of context is of vital importance for generation of accurate semantic information. In this research, a deep learning-based image parsing framework is proposed that utilizes novel relation-aware context learning technique. The proposed technique explores the graph constructs from the training data to learn the co-occurring context associations of object category labels using the graph edge connections. The proposed graph-based context learning technique defines the scene specific relation-awareness among semantic object categories, e.g., the probability of sky, road and building to co-exist in a scene is high. The proposed image parsing architecture (including the novel graph-based context learning technique) is evaluated on the benchmark datasets. In addition, a comprehensive comparison with existing image parsing techniques is presented to establish the efficacy of the scene-graph generation. The in-depth investigation of graph generation is presented to demonstrate the improvement in pixel-wise labeling.

Deep-learning models are popular machine learning models that have gained their popularity in various fields of computer vision, natural language processing etc, due to their excellent predictive accuracy and ability to automatically extract good features from raw input data. Though these models have these significant advantages, most deep learning models are notoriously black-box models. The features learned by these models are not explainable. The need to explain these predictions is important in many high stakes fields. In this work, we propose a method that uses convolutional layers like in Convolutional Neural Networks (CNNs) that automatically learns features from raw data. In our proposed method we extract feature representation that is unique to each target class in the given data. Visualizations of this unique representation explain the feature learned by the model for a specific target class. To classify an input data we use two approaches. The first approach uses the similarity scores to match the features learned for a particular instance with the extracted target class-specific features to make prediction. The second approach uses an intrinsically explainable model like logistic regression to make predictions. Thus our proposed method in addition to having good predictive accuracy, identifies features that are class-specific. The proposed method achieved an accuracy of 99.31 % on MNIST data set, 90.45% on CIFAR-10 data set and 96.45% accuracy on Assira data set.

Deep learning models have been very successful in computer vision and image processing applications. Since its inception, Convolutional Neural Network (CNN)-based deep learning models have consistently outperformed other machine learning methods on many significant image processing benchmarks. Many top-performing methods for image segmentation are also based on deep CNN models. However, deep CNN models fail to integrate global and local context alongside visual features despite having complex multi-layer architectures. We propose a novel three-layered deep learning model that learns independently global and local contextual information alongside visual features, and visual feature selection based on a genetic algorithm. The novelty of the proposed model is that One-vs-All binary class-based learners are introduced to learn Genetic Algorithm (GA) optimized features in the visual layer, followed by the contextual layer that learns global and local contexts of an image, and finally the third layer integrates all the information optimally to obtain the final class label. Stanford Background and CamVid benchmark image parsing datasets were used for our model evaluation, and our model shows promising results. The empirical analysis reveals that optimized visual features with global and local contextual information play a significant role to improve accuracy and produce stable predictions comparable to state-of-the-art deep CNN models.

Automatic interpretation of information written on roadside signs is very useful in applications for conducting road surveys, driving autonomous vehicles, and improving the road safety and infrastructure. However, it is a very difficult problem because of high similarity and poor visibility of signs and challenging nature of road environment. In this paper, we propose a novel technique for automatic detection and recognition of speed limit signs. The proposed technique contains two novel concepts. Firstly, pixel-wise segmentation of all speed limit signs into one class instead of each sign in a separate class is proposed. Secondly, a novel classifier with a false alarm class that evolves its weights so that it can distinguish speed signs from non-speed signs is proposed. The proposed technique is evaluated on a real-world dataset provided by our industry partners. The dataset has been created from videos of all state roads in Queensland. The comparative analysis of results showed that the proposed technique is able to detect and classify speed limit signs with high accuracy.

Feature selection for deep learning architectures is one of the important and challenging steps in developing an efficient image parsing application. In this paper, a novel image parsing architecture which makes use of unique feature selection is proposed. It introduces the idea of weighted relationship awareness to reduce the redundancy of features and optimally select an efficient subset of feature representations. The proposed architecture is evaluated on Cam Vid benchmark dataset. A comparison with state-of-the-art methods was conducted which showed significant improvements in terms of segmentation and classification accuracy.

Image classification is a fundamental task that attempts to apprehend an entire image by assigning a specific class or label. Accurate feature extraction techniques are the main building blocks of any image classification and prediction application. Most of these feature extraction techniques are manual (hand-crafted), hence they contain tedious, challenging, and erroneous tasks. In this paper, a new automatic technique for image classification is presented in which a novel concept to extract auto-associative features along with a non-iterative learning is investigated. The proposed technique incorporates an auto-associative neural network for most accurate feature extraction and a noniterative classifier. The proposed technique has been evaluated on benchmark datasets. The experimental results have demonstrated that the proposed technique can achieve same or slightly higher accuracy than the standard and formula-based techniques.

Ensemble classifiers using clustering have significantly improved classification and prediction accuracies of many systems. These types of ensemble approaches create multiple clusters to train the base classifiers. However, the problem with this is that each class might have many clusters and each cluster might have different number of samples, so an ensemble decision based on large number of clusters and different number of samples per class within a cluster produces biased and inaccurate results. Therefore, in this article, we propose a novel methodology to create an appropriate number of strong data clusters for each class and then balance them. Furthermore, an ensemble framework is proposed with base classifiers trained on strong and balanced data clusters. The proposed approach is implemented and evaluated on 24 benchmark data sets from the University of California Irvine (UCI) machine learning repository. An analysis of results using the proposed approach and the existing state-of-the-art ensemble classifier approaches is conducted and presented. A significance test is conducted to further validate the efficacy of the results and a detailed analysis is presented.

Clustering based ensemble classifiers have seen a lot of focus recently because of their ability to effectively classify real-world noisy datasets. One way of incorporating clustering in ensembles is to utilize clustering algorithms such as k-means to generate a pool of data clusters. This is done to generate a random subspace on which base classifiers are trained, as opposed to bagging. One key parameter to clustering algorithms is the number of clusters i.e. the number of groups the data should be partitioned into; this is commonly known as the variable K. Most of the existing approaches either determine the value of K through trial and error or use some derived formulae-based approach. The problem firstly is that using a static value of K for different datasets is not ideal and although a certain value may work well for one dataset it may not work well for others. Secondly, calculating the value of K using a formulae-based approach using the raw data is not effective either as an unbalanced data can have a negative effect on the derived value. Therefore, in this paper we first segregate the data based on the data classes and then on each data class we perform a Silhouette analysis to determine the optimal number of clusters each data class should be separated into. The generated clusters which are class pure and are balanced by adding samples from other classes that are closest to the cluster centroid. In this manner we generate a random subspace of an augmented data that is composed of class balanced data clusters. On all balanced data clusters, a diverse set of base classifiers is trained, and an ensemble is formed. The proposed ensemble approach is tested on 16 benchmark UCI datasets and results are compared with single classifiers, as well as state-of-the-art ensemble classifier approaches. A set of non-parametric tests are also adopted to further validate the efficacy of the results.

In the past few years, Convolutional Neural Networks (CNNs) have achieved surprisingly good results for objects classification in real world images. However, training a CNN from scratch for large datasets is still a nightmare, when it comes to time and resources. The main reason for this problem is long iterative training process used in CNN's fully connected layer which is also called a classification layer. Therefore, in this paper we propose a novel approach to make the convolutional neural network quicker and more efficient for image classification tasks. The proposed approach consists of a convolutional feature extraction layer and a non-iterative radial basis function-based classification layer. The proposed approach has been evaluated on three benchmark datasets such as CIFAR-10, MNIST and Digit. The experimental results have demonstrated that the proposed approach can achieve same or higher accuracy in lesser time than the standard CNN.

Deep Convolutional Neural Networks (DCNNs) have enhanced the performance of semantic image segmentation but many challenges still remain. Specifically, some details may be lost due to the downsampling operations in DCNNs. Furthermore, objects may appear in an image at different scales, and extracting features using convolutional filters with large sizes is costly in computation. Moreover, in many cases, contextual information, such as global and background features, is potentially useful for semantic segmentation. In this paper, we address these challenges by proposing a Multi-Receptive Atrous Convolutional Network (MRACN) for semantic image segmentation. The proposed MRACN captures the multi-receptive features and the global features at different receptive scales of the input. MRACN can serve as a module easily being integrated into existing models. We adapt the ResNet-101 model as the backbone network and further propose a MRACN segmentation model (MRACN-Seg). The experimental results demonstrate the effectiveness of the proposed model on two datasets: a benchmark dataset (PASCAL VOC 2012) and our industry dataset.

When generating ensemble classifiers, selecting the best set of classifiers from the base classifier pool is considered a combinatorial problem and an efficient classifier selection methodology must be utilized. Different researchers have used different strategies such as evolutionary algorithms, genetic algorithms, rule-based algorithms, simulated annealing, and so forth to select the best set of classifiers that can maximize overall ensemble classifier accuracy. In this article, we present a novel classifier selection approach to generate an ensemble classifier. The proposed approach selects classifiers in multiple rounds of elimination. In each round, a classifier is given a chance to be selected to become a part of the ensemble, if it can contribute to the overall ensemble accuracy or diversity; otherwise, it is put back into the pool. Each classifier is given multiple opportunities to participate in rounds of selection and they are discarded only if they have no remaining chances. The process is repeated until no classifier in the pool has any chance left to participate in the round of selection. To test the efficacy of the proposed approach, 13 benchmark datasets from the UCI repository are used and results are compared with single classifier models and existing state-of-the-art ensemble classifier approaches. Statistical significance testing is conducted to further validate the results, and an analysis is provided.

Ensemble classifiers improve the classification performance by combining several classifiers using a suitable fusion methodology. Many ensemble classifier generation methods have been developed that allowed the training of multiple classifiers on a single dataset. As such random subspace is a common methodology utilized by many state-of-the-art ensemble classifiers that generate random subsamples from the input data and train classifiers on different subsamples. Real-world datasets have randomness and noise in them, therefore not all randomly generated samples are suitable for training. In this article, we propose a novel particle swarm optimization-based approach to optimize the random subspace to generate an ensemble classifier. We first generate a random subspace by incrementally clustering input data and then optimize all generated data clusters. On all optimized data clusters, a set of classifiers is trained and added to the pool. The pool of classifiers is then optimized and an optimized ensemble classifier is generated. The proposed approach is tested on 12 benchmark datasets from the UCI repository and results are compared with current state-of-the-art ensemble classifier approaches. A statistical significance test is also conducted and an analysis is presented.

This paper proposes a new approach for evolving One-Dimensional Convolutional Neural Network (1D-CNN). A Particle Swarm Optimization (PSO) algorithm was incorporated to evolve the network layers and parameters. The proposed approach was evaluated over a real-world rainfall prediction application. It was compared to a trial and error approach and to the first version of the official rainfall prediction system used by the Bureau of Meteorology in Australia. Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and Pearson correlation (r) statistical measurements were calculated to measure the performance of each approach. The performance of the proposed approach was promising compared to existing state-of-the art methods. This proposed framework can be easily extended and deployed to other applications including machine vision.

Ensemble classifier approaches either exploit the input feature space also known as the dataset attributes, or exploit the data sample space, for example Random Forest (RaF) exploits input features whereas Bagging exploits data sample space. Very few ensemble classifier approaches exist that exploit the both. In this paper we propose an ensemble classifier approach that first reduces input feature space by selecting only the significant features from input data that can maximize the classification performance of the ensemble and then optimize the base classifier pool by incorporating an evolutionary algorithm. The proposed approach is evaluated on benchmark datasets from UCI repository. The results are compared with single classifier approaches and existing state of the art ensemble classifier approaches.

Diversity is a key component for building a successful ensemble classifier. One approach to diversifying the base classifiers in an ensemble classifier is to diversify the data they are trained on. While sampling approaches such as bagging have been used for this task in the past, we argue that since they maintain the global distribution, they do not create diversity. Instead, we make a principled argument for the use of [Formula: see text]-means clustering to create diversity. Expanding on previous work, we observe that when creating multiple clusterings with multiple [Formula: see text] values, there is a risk of different clusterings discovering the same clusters, which would in turn train the same base classifiers. This would bias the ensemble voting process. We propose a new approach that uses the Jaccard Index to detect and remove similar clusters before training the base classifiers, not only saving computation time, but also reducing classification error by removing repeated votes. We empirically demonstrate the effectiveness of the proposed approach compared to the state of the art on 19 UCI benchmark datasets.

Precipitation is viewed as a complex phenomenon that influences the efficiency of the agricultural season. In this paper, an ensemble of neural networks has been created and optimized to estimate monthly rainfall for Innisfail, Australia. The proposed ensemble utilizes single neural networks as components and combines them using a neural network fusion method. A novel ensemble components selection approach has been proposed and deployed. Ensemble components were selected based on a hybrid Genetic Algorithm (GA) that combines standard GA with particle swarm optimization algorithm. Various statistical measurements were calculated to assess the accuracy of the proposed ensembles against single neural networks, climatology and ensembles generated through an alternative selection approach. A better performance was obtained with the proposed ensembles when compared to alternative models.

Convolutional Neural Networks (CNNs) have demonstrated great potential in complex image classification problems in past few years. CNNs have a large number of parameters and the system accuracy depends directly on the selection of these parameters. With diverse parameters, selection of optimal parameter remains a trial and error, ad hoc or expert's mercy. In practice, optimal parameter selection remains the biggest obstacle in designing a real-world application using CNN. Convolutional neural network's performance is highly affected by its parameters. A novel approach is proposed in this paper to select convolutional neural network parameters in an image classification task. The proposed approach incorporated particle swarm optimization to select the parameters of the convolutional network. Two datasets, one benchmark CIFAR-IO and one real world application dataset, road-side vegetation dataset, were selected to evaluate the proposed approach. It is demonstrated that proposed approach efficiently explores the solution space, and determines the best combination of parameters. Extensive experiments, along with the statistical tests, revealed that proposed approach is an effective technique for automatically optimizing CNN's parameters.

We propose an ensemble classification algorithm that uses multi-objective optimization instead of relying on heuristics or fragile user-defined parameters. Only two user-defined parameters are included, with both being found to have large windows of values that produce statistically indistinguishable results, indicating the low level of expertise required from the user to achieve good results. Additionally, when given a large initial set of trained base-classifiers, we demonstrate that a multi-objective genetic algorithm aiming to optimize prediction accuracy and diversity will prefer particular types of classifiers over others. The total number of chosen classifiers is also surprisingly small - only 5.86 classifiers on average, out of a starting pool of 900. This occurs without any explicit preference for small ensembles of classifiers. Even with these small ensembles, significantly lower empirical classification error is achieved compared to the current state-of-the-art.

Divide-and-conquer-based methods are quite successful across various problems from different disciplines. These methods divide a complex task into multiple simple tasks and solve them collectively. This paper presents a divide-and-conquer-based hierarchical optimization framework for ensemble classifier learning (ECL). The optimization framework includes a search space creation process [called data training environments (DTE)] that divides the data into multiple clusters, and then trains a set of heterogeneous base classifiers with the DTEs. The classifiers are then combined to form an optimal ensemble, by finding the fittest ones using many-objective optimization. The many-objective optimization algorithm considers each class accuracy as a separate objective and maximizes the class accuracies. An additional objective is also taken into account by maximizing the ensemble size. Since the partitioning of data creates diversity within the pool of classifiers, class accuracy tradeoff among the classifiers is observed. As a result, increasing the number of classifiers also increases the diversity within the ensemble. In order to tackle the optimization, a specialized many-objective optimization algorithm based on decomposition is proposed. Since ECL can be regarded as an NP-hard problem, the proposed optimization algorithm, instead identifies the optimal ensemble using a divide-and-conquer rule-based chromosome encoding. Moreover, with the involvement of individual class accuracy in the objectives, the performance does not get biased toward any majority class. The proposed framework is experimented with 24 benchmark datasets obtained from the UCI machine learning repository and compared with the existing approaches. The experimental results show better classification accuracy with the proposed framework in comparison with the recent ensemble classifiers.

In this paper, we propose an incremental ensemble classifier learning method. In the proposed method, a set of accurate and diverse classifiers are generated and added to the ensemble by means of accuracy and diversity comparison. The selection of classifiers in ensemble starts with a layer (where data is partitioned into any given number of clusters and fed to a set of base classifiers) and then continues to improve the bias-variance (i.e., accuracy and diversity). Optimal ensemble classifier selection is done through accuracy-precedence-diversity comparison, i.e., a model with better accuracy is preferred but in the case of models with the same accuracy, better diversity is preferred. The comparison is made on the class decomposed accuracies (i.e., all class accuracies are decomposed to a scalar value). A non-identical set of base classifiers is trained on the clusters of data in a layer and the center of the cluster is recorded as an identifier to the corresponding base classifiers set. Decisions from multiple base classifiers are fused to an ensemble class output using majority voting for each pattern and finally the decisions across multiple layers are combined using majority voting. The proposed method is evaluated on UCI benchmark datasets and compared with the recently proposed ensemble classifiers including the Bagging and Boosting. Through comparison, we demonstrate that the proposed method improves the performance of the base classifiers and performs better than the existing ensemble methods.

Semantic context is an important and useful cue for scene parsing in complicated natural images with a substantial amount of variations in objects and the environment. This paper proposes Spatially Constrained Location Prior (SCLP) for effective modelling of global and local semantic context in the scene in terms of inter-class spatial relationships. Unlike existing studies focusing on either relative or absolute location prior of objects, the SCLP effectively incorporates both relative and absolute location priors by calculating object co-occurrence frequencies in spatially constrained image blocks. The SCLP is general and can be used in conjunction with various visual feature-based prediction models, such as Artificial Neural Networks and Support Vector Machine (SVM), to enforce spatial contextual constraints on class labels. Using SVM classifiers and a linear regression model, we demonstrate that the incorporation of SCLP achieves superior performance compared to the state-of-the-art methods on the Stanford background and SIFT Flow datasets.

Roadside vegetation classification has recently attracted increasing attention, due to its significance in applications such as vegetation growth management and fire hazard identification. Existing studies primarily focus on learning visible feature based classifiers or invisible feature based thresholds, which often suffer from a generalization problem to new data. This paper proposes an approach that aggregates pixel-level supervised classification and cluster-level texton occurrence within a voting strategy over superpixels for vegetation classification, which takes into account both generic features in the training data and local characteristics in the testing data. Class-specific artificial neural networks are trained to predict class probabilities for all pixels, while a texton based adaptive K-means clustering process is introduced to group pixels into clusters and obtain texton occurrence. The pixel-level class probabilities and cluster-level texton occurrence are further integrated in superpixel-level voting to assign each superpixel to a class category. The proposed approach outperforms previous approaches on a roadside image dataset collected by the Department of Transport and Main Roads, Queensland, Australia, and achieves state-of-the-art performance using low-resolution images from the Croatia roadside grass dataset.

“Happy New Year!” At the beginning of 2016, I would like to take this opportunity to wish everyone a very happy, healthy, and prosperous new year! It is my great honor and privilege to serve as the Editor-in-Chief (EiC) of the IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS (TNNLS), and I am excited to write this Editorial to start a new journey with you all.

Texture classification is an essential task in computer vision that aims at grouping instances that have a similar repetitive pattern into one group. Detecting texture primitives can be used to discriminate between materials of different types. The process of detecting prominent features from the texture instances represents a cornerstone step in texture classification. Moreover, building a good model using a few training instances is difficult. In this study, a genetic programming (GP) descriptor is proposed for the task of multiclass texture classification. The proposed method synthesises a set of mathematical formulas relying on the raw pixel values and a sliding window of a predetermined size. Furthermore, only two instances per class are used to automatically evolve a descriptor that has the potential to effectively discriminate between instances of different textures using a simple instance-based classifier to perform the classification task. The performance of the proposed approach is examined using two widely-used data sets, and compared with two GP-based and nine well-known non-GP methods. Furthermore, three hand-crafted domain-expert designed feature extraction methods have been used with the non-GP methods to examine the effectiveness of the proposed method. The results show that the proposed method has significantly outperformed all these other methods on both data sets, and the new method evolves a descriptor that is capable of achieving significantly better performance compared to hand-crafted features.

Genetic Programming (GP) has been applied to a wide range of image analysis tasks including many real-world segmentation problems. This paper introduces a new biological application of detecting Phormidium algae in rivers of New Zealand using raw images captured from the air. In this paper, we propose a GP method to the task of algae detection. The proposed method synthesises a set of image operators and adopts a simple thresholding approach to segmenting an image into algae and non-algae regions. Furthermore, the introduced method operates directly on raw pixel values with no human assistance required. The method is tested across seven different images from different rivers. The results show good success on detecting areas of algae much more efficiently than traditional manual techniques. Furthermore, the result achieved by the proposed method is comparable to the hand-crafted ground truth with a F-measure fitness value of 0.64 (where 0 is best, 1 is worst) on average on the test set. Issues such as illumination, reflection and waves are discussed.

Detecting and recognizing road objects automatically is an important process in many applications such as traffic regulation and providing guidance for drivers and pedestrians. Fuzzy clustering using wavelets is proposed in this paper. Wavelets are used for pre-processing the image and the resulting image is then subjected to fuzzy c-means algorithm for clustering. After clustering, the image classification is done by an ensemble of multi-layer perceptron neural networks. This approach is used to classify road images into different road side objects like road, sky, and signs. A database using real-world roadside images from Transport and Main Roads (TMR) is used for evaluating the proposed approach. The results on the database using the proposed approach indicate that this approach using wavelets improves the recognition rate. This approach is compared with existing methods for segmentation and classification of road images.

Classification of roadside objects is very important task in identifying fire risk regions, analysing roadside conditions and improving roadside safety. This paper introduces a novel and effective way to detect soil, grass, road and tree from roadside images thus giving a better decision-making system for analysing roadside video data. A new feature extraction approach is proposed to detect and classify the roadside objects. Feature set is based on colour characteristics which are obtained by analysing components of image pixels. Choosing an appropriate feature set is one of the great challenges for successful identification of roadside objects. Based on the proposed feature set and the Support Vector Machine, the detection and classification approach is implemented. The proposed approach is evaluated using the training and test data from real-world roadside video images. The results show that the proposed approach is able to accurately detect grass, soil, road and tree.

Determining the best values for the parameters of a classifier is a challenge. This challenge is compounded for ensembles. This research evaluates the number of neurons for candidate networks and the number of committee members in our work on variable neural classifiers for breast cancer diagnosis. The evaluation reveals that good neural network accuracy can be achieved with a small number of neurons in the hidden layer and three committee members in the ensemble. The proposed methodology is tested on two benchmark databases achieving 99% classification accuracy.

Ensemble classifiers are very useful tools which can be applied for classification and prediction tasks in many real-world applications. There are many popular ensemble classifier generation techniques including neural network based techniques. However, there are many problems with ensemble classifiers when we apply them to real-world data of different size. This paper presents and investigates an approach for finding the impact of various parameters such as attributes, instances, classes on clusters, accuracy and diversity. The primary aim of this research is to see whether there is any link between these parameters and accuracy and diversity. The secondary aim is to see whether we can find any relationship between number of clusters in ensemble classifier and data variables. A series of experiments has been conducted by using different size of UCI machine learning benchmark datasets and neural network ensemble classifiers.

In this paper, we present an application of Multi-Objective Evolutionary Algorithm (MOEA) for generating cluster oriented ensemble classifier. In our recently developed Non-Uniform Layered Cluster Oriented Ensemble Classifier (NULCOEC), the data set is partitioned into a variable number of clusters at different layers. Base classifiers are then trained on the clusters at different layers. The performance of NULCOEC is a function of the vector (layers, clusters) and the research presented in this paper investigates the implication of applying MOEA to generate NULCOEC. Accuracy and diversity of the ensemble classifier is expressed as a function of layers and clusters. A MOEA then searches for the combination of layers and clusters to obtain the non-dominated set of (accuracy, diversity). We have also obtained the results of single objective optimization (i.e. optimizing either accuracy or diversity) and compared them with the results of MOEA. The results show that the MOEA can improve the performance of ensemble classifier.

Abstract This paper presents cluster‐based ensemble classifier – an approach toward generating ensemble of classifiers using multiple clusters within classified data. Clustering is incorporated to partition data set into multiple clusters of highly correlated data that are difficult to separate otherwise and different base classifiers are used to learn class boundaries within the clusters. As the different base classifiers engage on different difficult‐to‐classify subsets of the data, the learning of the base classifiers is more focussed and accurate. A selection rather than fusion approach achieves the final verdict on patterns of unknown classes. The impact of clustering on the learning parameters and accuracy of a number of learning algorithms including neural network, support vector machine, decision tree and k‐NN classifier is investigated. A number of benchmark data sets from the UCI machine learning repository were used to evaluate the cluster‐based ensemble classifier and the experimental results demonstrate its superiority over bagging and boosting.

This paper presents an approach for analyzing relationships between data size, cluster, accuracy and diversity in neural network ensembles. The main objective of this research is to find out the influence of data size such as number of patterns, number of inputs and number of classes on various parameters such as clusters, accuracy and diversity of a neural network ensemble. The proposed approach is based on splitting data sets into different groups using the data size, clustering data and conducting training and testing of neural network ensembles. The test data is same for all groups and used to test all trained ensembles. The experiments have been conducted on 15 UCI machine learning benchmark datasets and results are presented in this paper.

This paper presents an algorithm to generate ensemble classifier by joint optimization of accuracy and diversity. It is expected that the base classifiers in an ensemble are accurate and diverse (i.e., complementary in terms of errors) among each other for the ensemble classifier to be more accurate. We adopt a multi-objective evolutionary algorithm (MOEA) for joint optimization of accuracy and diversity on our recently developed nonuniform layered cluster oriented ensemble classifier (NULCOEC). In NULCOEC, the data set is partitioned into a variable number of clusters at different layers. Base classifiers are then trained on the clusters at different layers. The performance of NULCOEC is a function of the vector of the number of layers and clusters. The research presented in this paper investigates the implication of applying MOEA to generate NULCOEC. Accuracy and diversity of the ensemble classifier is expressed as a function of layers and clusters. A MOEA then searches for the combination of layers and clusters to obtain the nondominated set of (accuracy, diversity). We have obtained the results of single objective optimization (i.e., optimizing either accuracy or diversity) and compared them with the results of MOEA on sixteen UCI data sets. The results show that the MOEA can improve the performance of ensemble classifier.

Neural networks provide an alternative approach for the solution of complex non-linear data fitting problems. In this paper, we propose a novel technique using a multilayer perceptron neural network to fit high energy stopping power data, where the unknown stopping power functional form was fitted to experimental data by a set of linear combination of neurons. The projectiles of Li, B, N, O, Ne and P in the solid matters C, Si, Ti and Ni are illustrated as examples of the application. Using the resilient backpropagation algorithm, it can obtain more accurate fitting coefficients than conventional iterative methods. Our simulations show that a simple, accurate predictor based on neural network fitting can produce reliable predictions of stopping power values either at the energy position or for the projectile-target combination where no measured data currently exist.

This paper proposes the creation of an ensemble neural network by incorporating a k-means classifier. This technique is designed to improve the classification accuracy of a multi-layer perceptron style network for mass classification of digital mammograms. The proposed technique has been tested on a benchmark database and the results have been contrasted with current research. The experimental results demonstrate that the accuracy of the proposed technique is comparable with existing systems.

In this paper, we have investigated the influence of cluster instability on the performance of layered cluster oriented ensemble classifier. The final contents of clusters in some clustering algorithms like k-means depend on the initialization of clustering parameters like cluster centres. Layered cluster oriented ensemble classifier is based on this philosophy where the base classifiers are trained on clusters generated at multiple layers from random initialization of cluster centres. As the data is clustered into multiple layers some patterns move between clusters (unstable patterns). This instability of patterns brings in diversity among the base classifiers that in turn influences the accuracy of the ensemble classifier. There is thus a connection between the instability of the patterns and the accuracy of layered cluster oriented ensemble classifier. The research presented in this paper aims to find this connection by investigating the influence of unstable patterns on the overall ensemble classifier accuracy as well as diversity among the base classifiers. We have provided results from a number of experiments to quantify this influence.

The segmentation of road images from vehicle mounted video is a challenging and difficult problem. One of the problems is the presence of different types of objects and not all objects are present in the same frame. For example, road sign is not visible in all frames. In this paper, we propose a novel framework for segmenting road images in a hierarchical manner that can separate the following objects: sky, road, road signs, and vegetation from the video data. Each frame in the video is analysed separately. The hierarchical approach does not assume the presence of a certain number of objects in a single frame. We have also developed a segmentation framework based on SVM learning. The proposed framework has been tested on the Transport and Main Roads Queensland's video data. The experimental results indicate that the proposed framework can detect different objects with an accuracy of 95.65%.

This paper presents a novel cluster-oriented ensemble classifier. The proposed ensemble classifier is based on original concepts such as learning of cluster boundaries by the base classifiers and mapping of cluster confidences to class decision using a fusion classifier. The categorized data set is characterized into multiple clusters and fed to a number of distinctive base classifiers. The base classifiers learn cluster boundaries and produce cluster confidence vectors. A second level fusion classifier combines the cluster confidences and maps to class decisions. The proposed ensemble classifier modifies the learning domain for the base classifiers and facilitates efficient learning. The proposed approach is evaluated on benchmark data sets from UCI machine learning repository to identify the impact of multicluster boundaries on classifier learning and classification accuracy. The experimental results and two-tailed sign test demonstrate the superiority of the proposed cluster-oriented ensemble classifier over existing ensemble classifiers published in the literature.

In this paper we propose different ensemble classifier compositions and investigate their influence on offline cursive character recognition. Cursive characters are difficult to recognize due to different handwriting styles of different writers. The recognition accuracy can be improved by training an ensemble of classifiers on multiple feature sets focussing on different aspects of character images. Given the feature sets and base classifiers, we have developed multiple ensemble classifier compositions using three architectures. Type-1 architecture is based on homogeneous base classifiers and Type-2 architecture is composed of heterogeneous base classifiers. Type-3 architecture is based on hierarchical fusion of decisions. The experimental results demonstrate that the presented method with best composition of classifiers and feature sets performs better than existing methods for offline cursive character recognition.

This paper presents a novel technique for the classification of suspicious areas in digital mammograms. The proposed technique is based on clustering of input data into numerous clusters and amalgamating them with a Support Vector Machine (SVM) classifier. The technique is called multi-cluster support vector machine (MCSVM) and is designed to provide a fast converging technique with good generalization abilities leading to an improved classification as a benign or malignant class. The proposed MCSVM technique has been evaluated on data from the Digital Database of Screening Mammography (DDSM) benchmark database. The experimental results showed that the proposed MCSVM classifier achieves better results than standard SVM. A paired t-test and Anova analysis showed that the results are statistically significant.

This paper introduces a novel concept for creating an ensemble of classifiers. The concept is based on generating an ensemble of classifiers through clustering of data at multiple layers. The ensemble classifier model generates a set of alternative clustering of a dataset at different layers by randomly initializing the clustering parameters and trains a set of base classifiers on the patterns at different clusters in different layers. A test pattern is classified by first finding the appropriate cluster at each layer and then using the corresponding base classifier. The decisions obtained at different layers are fused into a final verdict using majority voting. As the base classifiers are trained on overlapping patterns at different layers, the proposed approach achieves diversity among the individual classifiers. Identification of difficult-to-classify patterns through clustering as well as achievement of diversity through layering leads to better classification results as evidenced from the experimental results.

This paper presents an evolutionary algorithm based approach for selection of topologies in hierarchical fuzzy systems (HFS). The different topologies of HFS for a given dynamical system such as inverted pendulum are investigated and analysed to address the problem of how input configuration in multi-layered structure affects the controller performance. The experiments are conducted to test controller performance for different topologies of the hierarchical fuzzy system. The impact of different topologies on control process is discussed. The results from case study (inverted pendulum) can be extended to other dynamical systems, in particular to a robotic manipulator.

A compositional method for hierarchical fuzzy control over the user-defined set of initial conditions is presented. The control system is found by using a multi-objective evolutionary algorithm. The inverted pendulum system is selected as an example of a dynamical system and used to test the proposed method. The pre-defined set of initial conditions includes dynamical and static conditions of the system. Control system is designed as a three-layered hierarchical fuzzy logic structure. The test results showed that the proposed method has a relatively high success rate in terms of the number of initial conditions from which the system is controlled to TR. The best achieved result was 94.5% success rate. The proposed compositional method can be applied to a wide range of dynamical systems.

A novel Over-Segmentation and Neural Binary Validation (OSNBV) is presented in this paper. OSNBV is a character segmentation strategy for off-line cursive handwriting recognition. Unlike the approaches in the literature, OSNBV is a prioritized segmentation approach. Initially, OSNBV over-segments a handwritten word into primitives. Neural binary validation is iteratively applied to the primitives. The outcome of each iteration is to join two neighboring primitives when the joined one improves the global neural competency. OSNBV introduces Transition Count (TC) and TC for English (EngTC) to prevent under-segmentation error during neural binary validation. OSNBV also incorporates Transition Count Matrix (TCM) into neural global competency. The proposed approach has been evaluated on CEDAR benchmark database. The results showed a significant improvement in segmentation errors. The analysis of results showed that the inclusion of TCM into the validation function has played a major role in improving over-segmentation and bad-segmentation errors.

This paper presents a novel methodology for the classification of suspicious areas in digital mammograms. The methodology is based on the fusion of clustered sub classes with various intelligent classifiers. A number of classifiers have been incorporated into the proposed methodology and evaluated on the well known benchmark digital database of screening mammography (DDSM). The results in the form of overall classification accuracies, TP, TN, FP and FN have been analyzed, compared and presented. The results of all four tested classifiers with clustered sub classes on the DDSM benchmark database show that the proposed methodology can significantly improve the accuracy and reduce the false positive rate.

This paper presents a novel approach for creating and training of an ensemble classifier. The approach is based on creating atomic and non-atomic clusters at different levels, training of weak classifiers on overlapping clusters and fusion of their decisions. The subsets of data are obtained by clustering of original training data sets into multiple partitions. As each partition represents highly correlated patterns from different classes, the proposed approach trains weak classifiers on difficult-to-classify patterns and combines the decision at various levels. The approach is tested on six benchmark datasets from UCI machine learning repository. The results show that the proposed approach achieves better classification accuracy than the existing approaches.

Over-Segmentation and validation (OSV) is a well anticipated segmentation strategy in cursive off-line handwriting recognition. Over-Segmentation is a means of locating all possible character boundaries, and the excessive segmentation points called over-segmentation points. Validation is a process to check and validate the segmentation points whether or not they are correct character boundaries by commonly employing an intelligent classifier trained with knowledge of characters. The existing OSV algorithms use ordered validation which means that the incorrect segmentation points might account for the validity of the next segmentation point. The ordered validation creates problems such as chain-failure. This paper presents a novel Binary Segmentation with Neural Validation (BSNV) to reduce the chain-failure. BSNV contains modules of over-segmentation and validation but the main distinctive feature of BSNV is an un-ordered segmentation strategy. The proposed algorithm has been evaluated on CEDAR benchmark database and the results of the experiments are promising.

This paper evaluates the impact of multiple clusters on neural classification of regions of interest (ROIs) in digital mammograms. The training and test sets for neural networks usually contain inputs extracted from ROIs and relevant class such as benign and malignant. However, the patterns such as regions of interest in digital mammograms do not have just one cluster per class instead they have many clusters within benign and malignant classes. Therefore, neural network training may benefit in terms of accuracy and efficiency by creating and analyzing a number of clusters within a class. A novel multiple clusters based neural classification approach is presented. In this approach, input data is clustered into a number of clusters per class and a neural classifier is trained with clustered data which contain multiple clusters per class. The experiments on a benchmark database of digital mammograms are conducted. The results show that the multiple clusters per class have significant impact on neural classification and overall they achieve better accuracy than single cluster per class based classification of ROIs in digital mammograms.

This paper presents a novel compositional method for finding fuzzy rules in a three-layered hierarchical fuzzy structure. The proposed method incorporates a multi-objective evolutionary algorithm and a large set of initial conditions, including dynamical conditions of the system under investigation. The proposed method is focused on handling the large set of initial conditions by a multi-objective evolutionary algorithm and it can be applied to a wide range of dynamical control systems in robotics. The method has been evaluated on a dynamical system such as the inverted pendulum. The experimental results and analysis showed that the proposed method is much better than the existing methods such as amalgamation and single objective evolutionary algorithm based methods.

This paper presents a novel segmentation algorithm for offline cursive handwriting recognition. An over-segmentation algorithm is introduced to dissect the words from handwritten text based on the pixel density between upper and lower baselines. Each segment from the over-segmentation is passed to a multiple expert-based validation process. First expert compares the total foreground pixel of the segmentation point to a threshold value. The threshold is set and calculated before the segmentation by scanning the stroke components in the word. Second expert checks for closed areas such as holes. Third expert validates segmentation points using a neural voting approach which is trained on segmented characters before validation process starts. Final expert is based on oversized segment analysis to detect possible missed segmentation points. The proposed algorithm has been implemented and the experiments on cursive handwritten text have been conducted. The results of the experiments are very promising and the overall performance of the algorithm is more effective than the other existing segmentation algorithms.

The paper presents a segmentation based adaptive approach for the learning and recognition of single person's handwritten text. The approach is incorporated into an automated intelligent system for scanning of handwritten text on a paper and converting it into a text file. It scans an A4 size handwritten page and segments it into lines, words and characters. The segmented characters are passed to a neural classifier for the recognition. The final word is passed through a lexicon based matching process to improve the accuracy of the recognized text. Two neural networks are investigated for the learning of segmented characters quickly and accurately. The experimental results show that the proposed approach can produce high text recognition accuracy with a small number of training samples.

This paper presents a novel hybrid data mining approach for knowledge extraction and classification in medical databases. The approach combines self organizing map, k-means and naive Bayes with a neural network based classifier. The idea is to cluster all data in soft clusters using neural and statistical clustering and fuse them using serial and parallel fusion in conjunction with a neural classifier. The approach has been implemented and tested on a benchmark medical database. The preliminary experiments are very promising.

This paper presents a neural network based technique for the classification of segments of road images into cracks and normal images. The density and histogram features are extracted. The features are passed to a neural network for the classification of images into images with and without cracks. Once images are classified into cracks and non-cracks, they are passed to another neural network for the classification of a crack type after segmentation. Some experiments were conducted and promising results were obtained. The selected results and a comparative analysis are included in this paper.

This paper presents a new learning algorithm for the diagnosis of breast cancer. The proposed algorithm with novel network architecture can memorize training patterns with 100% retrieval accuracy as well as achieve high generalization accuracy for patterns which it has never seen before. The grey-level and BI-RADS features (radiologists' interpretation) from digital mammograms are extracted and used to train the network with the proposed learning algorithm. The new learning algorithm has been implemented and tested on a DDSM Benchmark database. The proposed approach has outperformed other existing approaches in terms of classification rate, generalization and memorization abilities, number of iterations, fast and guaranteed training. Some promising results and a comparative analysis of obtained results are included in this paper.

Early detection of breast abnormalities remains the primary prevention against breast cancer despite the advances in breast cancer diagnosis and treatment. Presence of mass in breast tissues is highly indicative of breast cancer. The research work presented in this paper investigates the significance of different types of features using proposed neural network based classification technique to classify mass type of breast abnormalities in digital mammograms into malignant and benign. 14 gray level based features, four BI-RADS features, patient age feature and subtlety value feature have been explored using the proposed research methodology to attain maximum classification on test dataset. The proposed research technique attained a 91% testing classification rate with a 100% training classification rate on digital mammograms taken from the DDSM benchmark database.

In this paper we present a novel technique for determining term importance by exploiting concept-based information found in ontologies. Calculating term importance is a significant and fundamental aspect of most information retrieval approaches, and it is traditionally determined through inverse document frequency (IDF). We propose concept-based term weighting (CBW), a technique that is fundamentally different to IDF in that it calculates term importance by intuitively interpreting the conceptual information in ontologies. We show that when CBW is used in an approach for web information retrieval on benchmark data, it performs comparatively to IDF, with only a 3.5% degradation in retrieval accuracy. While this small degradation has been observed, the significance of this technique is that (1) unlike IDF, CBW is independent of document collection statistics, (2) it presents a new way of interpreting ontologies for retrieval, and (3) it introduces an additional source of term importance information that can be used for term weighting.

Breast cancer continues to be the most common cause of cancer deaths in women. Early detection of breast cancer is significant for better prognosis. Digital Mammography currently offers the best control strategy for the early detection of breast cancer. The research work in this paper investigates the significance of neural-association of microcalcification patterns for their reliable classification in digital mammograms. The proposed technique explores the auto-associative abilities of a neural network approach to regenerate the composite of its learned patterns most consistent with the new information, thus the regenerated patterns can uniquely signify each input class and improve the overall classification. Two types of features: computer extracted (gray level based statistical) features and human extracted (radiologists' interpretation) features are used for the classification of calcification type of breast abnormalities. The proposed technique attained the highest 90.5% classification rate on the calcification testing dataset.

This paper presents a novel context matching technique for the retrieval of Web documents. The aim of the technique is to dynamically generate a context-based measure of document term significance during retrieval that can be used as a substitute or co-contributor of the term frequency measure. Unlike term frequency, which relies on a term to occur multiple times within a document to be considered significant, context matching is based on the notion that if a term in a given document occurs in that document in the context of the query, then that term is deemed to be significant. Context matching has the ability to potentially determine a term to be significant even if it occurs only once in a large document. The proposed technique has been implemented and the experiments were conducted using a TREC benchmark database. A comparative analysis shows that context matching significantly improves retrieval effectiveness and outperforms previously published results.

This paper investigates the significance of combining grey-level based image features and BI-RADS lesion descriptors along with patient age and a subtlety value (radiologists' interpretation) for the reliable classification of calcification and mass type breast abnormalities into malignant and benign classes. Three sets of experiments using grey-level based image features, BI-RADS features and combined features were conducted on DDSIM benchmark database. The classification rate 91% on mass dataset and 74% on calcification dataset was obtained when both types of features combined together.

In this paper we present an approach based on the application of an evolutionary algorithm to optimally tune the parameters of a novel technique for effective Web information retrieval. Context matching is a context-based technique for the ad-hoc retrieval of Web documents that relies on a number of inter-related parameters that define the nature of the context it uses. Its aim is to dynamically generate a context-based measure of term significance during retrieval that can be used as an indicator of document relevancy and ultimately contribute to a documents rank score. But the optimal setting of context matching parameters is an important aspect of the technique to ensure effective retrieval. Thus, the goal of this paper is to investigate the use of an evolutionary algorithm for the optimization of context matching parameters and compare its performance to an iterative technique that exhaustively explores combinations of parameters. We show how the most effective settings for parameters are obtained efficiently through the evolutionary algorithm. We also show how context matching, through the use of these optimized parameters, achieves effective retrieval results on benchmark data that are a significant improvement on previously published results.

This paper describes a neural network-based technique for cursive character recognition applicable to segmentation-based word recognition systems. The proposed research builds on a novel feature extraction technique that extracts direction information from the structure of character contours. This principal is extended so that the direction information is integrated with a technique for detecting transitions between background and foreground pixels in the character image. The proposed technique is compared with the standard direction feature extraction technique, providing promising results using segmented characters from the CEDAR benchmark database.

The paper presents a feature extraction technique for online handwriting recognition. The technique incorporates many characteristics of handwritten characters based on structural, directional and zoning information and combines them to create a single global feature vector. The technique is independent to character size and it can extract features from the raw data without resizing. Using the proposed technique and a neural network based classifier, many experiments were conducted on UNIPEN benchmark database. The recognition rates are 98.2% for digits, 91.2% for uppercase and 91.4% for lowercase.

Digital mammography is one of the most suitable methods for early detection of breast cancer. In uses digital mammograms to find suspicious areas. However, it is very difficult to distinguish benign and malignant cases, especially for the small size lesions in the early stage of cancer. This is reflected in the high percentage of unnecessary biopsies that are performed and many deaths caused by late detection or misdiagnosis. A computer based feature selection and classification system can provide a second opinion to the radiologists. This work proposes a neural-genetic algorithm for feature selection in conjunction with neural network based classifier. It also combined the computer-extracted statistical features from the mammogram with the human-extracted features for classifying different types of small breast abnormalities. It obtained 90.5% accuracy rate for calcification cases and 87.2% for mass cases with difference feature subsets. The obtained results show that different types of breast abnormality should use different features for classification.

In this paper we propose a fully automated offline handwriting recognition system that incorporates rule based segmentation, contour based feature extraction, neural network validation, a hybrid neural network classifier and a hamming neural network lexicon. The work is based on our earlier promising results in this area using heuristic segmentation and contour based feature extraction. The segmentation is done using many heuristic based set of rules in an iterative manner and finally followed by a neural network validation system. The extraction of feature is performed using both contour and structure based feature extraction algorithm. The classification is performed by a hybrid neural network that incorporates a hybrid combination of evolutionary algorithm and matrix based solution method. Finally a hamming neural network is used as a lexicon. A benchmark dataset from CEDAR has been used for training and testing.

Digital mammography is one of the most suitable methods for early detection of breast cancer. It uses digital mammograms to find suspicious areas containing benign and malignant microcalcifications. However, it is very difficult to distinguish benign and malignant microcalcifications. This is reflected in the high percentage of unnecessary biopsies that are performed and many deaths caused by late detection or misdiagnosis. A computer based feature selection and classification system can provide a second opinion to the radiologists in assessment of microcalcifications. The research proposes and investigates a neural-genetic algorithm for feature selection in conjunction with neural and statistical classifiers to classify microcalcification patterns in digital mammograms. The obtained results show that the proposed approach is able to find an appropriate feature subset and neural classifier achieves better results than two statistical models.

High accuracy character recognition techniques can provide useful information for segmentation-based handwritten word recognition systems. This research describes neural network-based techniques for segmented character recognition that may be applied to the segmentation and recognition components of an off-line handwritten word recognition system. Two neural architectures along with two different feature extraction techniques were investigated. A novel technique for character feature extraction is discussed and compared with others in the literature. Recognition results above 80% are reported using characters automatically segmented from the CEDAR benchmark database as well as standard CEDAR alphanumerics.

The purpose of this paper is to present a novel contour code feature in conjunction with a rule based segmentation for cursive handwriting recognition. A heuristic segmentation algorithm is initially used to over segment each word. Then the prospective segmentation points are passed through the rule-based module to discard the incorrect segmentation points and include any missing segmentation points. The proposed rule-based module validates every segmentation points against closed area, average character size, left character and density. During the left char validation, a contour code feature is extracted and checked weather the left of the prospective segmentation point is a character or rubbish (non-char). The neural network used for this validation was trained on character and non-character database. Following the segmentation, the contour between correct segmentation points is passed through the feature extraction module that extracts the contour code, after which another trained neural network is used for classification. The recognized characters are grouped into words and passed to a variable length lexicon that retrieves words that has highest confidence value. 1.

An automatic recognition of online handwritten text has been an on-going research problem for nearly four decades. It has been gaining more interest due to the increasing popularity of hand-held computers, digital notebooks and advanced cellular phones. However for these input modalities to be economical and user friendly the recognition rate should be very high for real time use. Also, the large number of writing styles and the variability between them makes the handwriting recognition problem a very challenging area for researchers. Many researchers have proposed a number of novel techniques for online handwriting recognition. However, an acceptable classification rate has not been achieved yet and there is a lack of techniques, which can find appropriate features, architecture and network parameters for online handwriting recognition. In this paper we propose a novel neurogenetic technique to improve classification accuracy through the selection of appropriate features and network parameters for online handwriting recognition. The technique incorporates an evolutionary approach for finding the most significant features, network architecture and its parameters. 1.

The paper proposes a novel min-max feature value based neural architecture and learning algorithm for classification of microcalcification patterns in digital mammograms. The neural architecture has a single hidden layer and it has a fixed number of hidden units and outputs. One class is represented by three hidden units and an output. The suspicious areas represented by chain code, are extracted from digital mammograms. The feature values are extracted for benign and malignant microcalcifications. A set of min, average and max values for every input feature is defined and assigned to the weights between input and hidden layer. The weights of the output layer are calculated using least squares methods or assigned in such a way that it maximizes the output value for only one class. Many experiments were conducted on a benchmark database of digital mammograms and comparative results are included in this paper.

In this paper, we present a novel approach of implementing a combination methodology to find appropriate neural network architecture and weights using an evolutionary least square based algorithm (GALS).1 This paper focuses on aspects such as the heuristics of updating weights using an evolutionary least square based algorithm, finding the number of hidden neurons for a two layer feed forward neural network, the stopping criterion for the algorithm and finally some comparisons of the results with other existing methods for searching optimal or near optimal solution in the multidimensional complex search space comprising the architecture and the weight variables. We explain how the weight updating algorithm using evolutionary least square based approach can be combined with the growing architecture model to find the optimum number of hidden neurons. We also discuss the issues of finding a probabilistic solution space as a starting point for the least square method and address the problems involving fitness breaking. We apply the proposed approach to XOR problem, 10 bit odd parity problem and many real-world benchmark data sets such as handwriting data set from CEDAR, breast cancer and heart disease data sets from UCI ML repository. The comparative results based on classification accuracy and the time complexity are discussed.

We present a novel genetic algorithm and least square (GALS) based hybrid learning approach for the training of an artificial neural network (ANN). The approach combines evolutionary algorithms with matrix solution methods such as Gram-Schmidt, SVD, etc., to adjust weights for hidden and output layers. Our hybrid method (GALS) incorporates the evolutionary algorithm (EA) in the first layer and the least square method (LS) in the second layer of the ANN. In the proposed approach, a two-layer network is considered, the hidden layer weights are evolved using an evolutionary algorithm and the output layer weights are calculated using a linear least square method. When a certain number of generation or error goals in terms of RMS error is reached, the training is stopped. We start training with a small number of hidden neurons and then the number is increased gradually in an incremental process. The proposed algorithm was implemented and many experiments were conducted on benchmark data sets such as XOR, 10-bit odd parity, handwritten segmented characters recognition, breast cancer diagnosis and heart disease data. The experimental results showed very promising results when compared with other existing evolutionary and error back propagation (EBP) algorithms in classification rate and time complexity.

This paper presents a comparative analysis of segmentation and non-segmentation based techniques for cursive handwritten word recognition. In our segmentation based technique, every word is segmented into characters, the chain code features are extracted from segmented characters, the features are fed to neural network classifier and finally the words are constructed using a string compare algorithm. In our non-segmentation based technique, the chain code features are extracted directly from words and the words are fed to a neural network classifier to classify them into word classes. To make a fair comparison, a CEDAR benchmark database is used, and the parameters such as the number of words, thresholding, resizing, feature extraction techniques, etc. are kept same for both the techniques. The experimental results and analysis show that the non-segmentation technique achieves higher recognition rate than the segmentation based technique.

The paper presents an intelligent hybrid approach for content-based image retrieval based on texture feature. The proposed approach employs an Auto–Associative Neural Network (AANN) for feature extraction and a Multi–Layer Perceptron (MLP) with a single hidden layer for the classification. Two intelligent approaches such as AANN–MLP and statistical–MLP were investigated. The performance of the proposed approaches was evaluated on a large benchmark database of texture patterns. The results are very promising compared to other existing traditional and intelligent techniques. Some of the experimental results conducted during the investigation, comparative analysis of the results and suggestions to select the appropriate techniques for texture feature extraction and classification are presented in this paper.

Analyses the performance of our improved segmentation algorithm tested on the CEDAR benchmark database of handwritten words. Segmentation is achieved through the extraction of a wide range of information adjacent to or surrounding suspicious segmentation points. Initially, a heuristic technique is employed to search for structural features and to over-segment each word. For each segmentation point that is located, the left character (preceding the segmentation point) and centre character (centred on the segmentation point) are extracted along with other features from the segmentation area. The aforementioned features are presented to trained character and segmentation point validation neural networks to evaluate a number of confidence values. Finally, the confidence values are fused to obtain the final segmentation decision. Based on a detailed analysis, it was observed that the left and centre character networks increased the accuracy of the segmentation algorithm.

An intelligent computer-aided diagnosis system can be very helpful for radiologist in detecting and diagnosing microcalcifications' patterns earlier and faster than typical screening programs. In this paper, we present a system based on fuzzy-neural and feature extraction techniques for detecting and diagnosing microcalcifications' patterns in digital mammograms. We have investigated and analyzed a number of feature extraction techniques and found that a combination of three features, such as entropy, standard deviation, and number of pixels, is the best combination to distinguish a benign microcalcification pattern from one that is malignant. A fuzzy technique in conjunction with three features was used to detect a microcalcification pattern and a neural network to classify it into benign/malignant. The system was developed on a Windows platform. It is an easy to use intelligent system that gives the user options to diagnose, detect, enlarge, zoom, and measure distances of areas in digital mammograms.

A new intelligent segmentation technique is proposed that may be used in conjunction with a neural classifier and a simple lexicon for the recognition of difficult handwritten words. A heuristic segmentation algorithm is initially used to over-segment each word. An artificial neural network (ANN) trained with 32,034 segmentation points is then used to verify the validity of the segmentation points found. Following segmentation, character matrices from each word are extracted, normalised and then passed through a global feature extractor, after which a second ANN trained with segmented characters is used for classification. These recognised characters are grouped into words and presented to a variable-length lexicon that utilises a string processing algorithm to compare and retrieve those words with the highest confidences. This research provides promising results for segmentation, character and word recognition.

An algorithm for segmenting unconstrained printed and cursive words is proposed. The algorithm initially oversegments handwritten word images (for training and testing) using heuristics and feature detection. An artificial neural network (ANN) is then trained with global features extracted from segmentation points found in words designated for training. Segmentation points located in "test" word images are subsequently extracted and verified using the trained ANN. Two major sets of experiments were conducted, resulting in segmentation accuracies of 75.06% and 76.52%. The handwritten words used for experimentation were taken from the CEDAR CD-ROM. The results obtained for segmentation can easily be used for comparison with other researchers using the same benchmark database.

In this paper, we propose a neuro-fuzzy technique for content based image retrieval. The technique is based on fuzzy interpretation of natural language, neural network learning and searching algorithms. Firstly, fuzzy logic is developed to interpret natural expressions such as mostly, many and few. Secondly, a neural network is designed to learn the meaning of mostly red, many red and few red. The neural network is independent to the database used, which avoids re-training of the neural network. Finally, a binary search algorithm is used to match and display neural network's output and images from database. The proposed technique is very unique and the originality of this research is not only based on hybrid approach to content based image retrieval but also on the new idea of training neural networks on queries. One of the most unique aspects of this research is that neural network is designed to learn queries and not databases. The technique can be used for any real-world online database. The technique has been implemented using CGI scripts and C programming language. Experimental results demonstrate the success of the new approach.

This paper presents a new automatic feature extraction technique and a neural network based classification method for recognition of rotating images. The image processing technique extracts global features of an image and converts a large size image into a one-dimensional small vector. A special advantage of the proposed technique is that the extracted features are the same even if the original image is rotated with rotation angles from 5 to 355 or rotated and a little bit distorted. The proposed technique is based on simple co-ordinate geometry, fuzzy sets and neural networks. The proposed approach is very easy in implementation and it has been developed in C++ on a Sun workstation. The experimental results have demonstrated that the proposed approach performs successfully on a variety of small as well as large scale rotated and distorted images.

Training a multilayer perceptron by an error backpropagation algorithm is slow and uncertain. This paper describes a new approach which is much faster and certain than error backpropagation. The proposed approach is based on combined iterative and direct solution methods. In this approach, we use an inverse transformation for linearization of nonlinear output activation functions, direct solution matrix methods for training the weights of the output layer; and gradient descent, the delta rule, and other proposed techniques for training the weights of the hidden layers. The approach has been implemented and tested on many problems. Experimental results, including training times and recognition accuracy, are given. Generally, the approach achieves accuracy as good as or better than perceptrons trained using error backpropagation, and the training process is much faster than the error backpropagation algorithm and also avoids local minima and paralysis.