A Backward Feature Selection by Creating Compact Neural Network Using Coherence Learning and Pruning

In this paper we propose a new backward feature selection method that generates compact classifier of a three-layered feed-forward artificial neural network (ANN). In the algorithm, that is based on the wrapper model, two techniques, coherence and pruning, are integrated together in order to find relevant features with a network of minimal numbers of hidden units and connections. Firstly, a coherence learning and a pruning technique are applied during training for removing unnecessary hidden units from the network. After that, attribute distances are measured by a straightforward computation that is not computationally expensive. An attribute is then removed based on an error-based criterion. The network is retrained after the removal of the attribute. This unnecessary attribute selection process is continued until a stopping criterion is satisfied. We applied this method to several standard benchmark classification problems such as breast cancer, diabetes, glass identification and thyroid problems. Experimental results confirmed that the proposed method generates compact network structures that can select relevant features with good classification accuracies.

“Hippocrates-mst”: a prototype for computer-aided microcalcification analysis and risk assessment for breast cancer

One of the most common cancer types among women is breast cancer. Regular mammographic examinations increase the possibility for early diagnosis and treatment and significantly improve the chance of survival for patients with breast cancer. Clustered microcalcifications have been considered as important indicators of the presence of breast cancer. We present "Hippocrates-mst", a prototype system for computer-aided risk assessment of breast cancer. Our research has been focused in developing software to locate microcalcifications on X-ray mammography images, quantify their critical features and classify them according to their probability of being cancerous. A total of 260 cases (187 benign and 73 malignant) have been examined and the performance of the prototype is presented through receiver operating characteristic (ROC) analysis. The system is showing high levels of sensitivity identifying correctly 98.63% of malignant cases.

Supervisory control of mobile sensor networks: math formulation, simulation, and implementation

This paper uses a novel discrete-event controller (DEC) for the coordination of cooperating heterogeneous wireless sensor networks (WSNs) containing both unattended ground sensors (UGSs) and mobile sensor robots. The DEC sequences the most suitable tasks for each agent and assigns sensor resources according to the current perception of the environment. A matrix formulation makes this DEC particularly useful for WSN, where missions change and sensor agents may be added or may fail. WSN have peculiarities that complicate their supervisory control. Therefore, this paper introduces several new tools for DEC design and operation, including methods for generating the required supervisory matrices based on mission planning, methods for modifying the matrices in the event of failed nodes, or nodes entering the network, and a novel dynamic priority assignment weighting approach for selecting the most appropriate and useful sensors for a given mission task. The resulting DEC represents a complete dynamical description of the WSN system, which allows a fast programming of deployable WSN, a computer simulation analysis, and an efficient implementation. The DEC is actually implemented on an experimental wireless-sensor-network prototyping system. Both simulation and experimental results are presented to show the effectiveness and versatility of the developed control architecture.

A NEW TECHNIQUE FOR SELECTING FEATURES FROM PROTEIN SEQUENCES

A new method for selecting features from protein sequences is proposed in this paper. First, the protein sequences are converted into fixed-dimensional feature vectors. Then, a subset of features is selected using relative entropy method and used as the inputs for Support Vector Machine (SVM). Finally, the trained SVM classifier is utilized to classify protein sequences into certain known protein families. Experimental results over proteins obtained from PIR database and GPCRs have shown that our proposed approach is really effective and efficient in selecting features from protein sequences.

Global model analysis by parameter space partitioning

To model behavior, scientists need to know how models behave. This means learning what other behaviors a model can produce besides the one generated by participants in an experiment. This is a difficult problem because of the complexity of psychological models (e.g., their many parameters) and because the behavioral precision of models (e.g., interval-scale performance) often mismatches their testable precision in experiments, where qualitative, ordinal predictions are the norm. Parameter space partitioning is a solution that evaluates model performance at a qualitative level. There exists a partition on the model's parameter space that divides it into regions that correspond to each data pattern. Three application examples demonstrate its potential and versatility for studying the global behavior of psychological models.

Application of artificial neural networks for gradient elution retention modelling in ion chromatography

Gradient elution in ion chromatography (IC) offers several advantages: total analysis time can be significantly reduced, overall resolution of a mixture can be increased, peak shape can be improved (less tailing) and effective sensitivity can be increased (because there is little variation in peak shape). More importantly, it provides the maximum resolution per time unit. The aim of this work was the development of a suitable artificial neural network (ANN) gradient elution retention model that can be used in a variety of applications for method development and retention modelling of inorganic anions in IC. Multilayer perceptron ANNs were used to model the retention behaviour of fluoride, chloride, nitrite, sulphate, bromide, nitrate and phosphate in relation to the starting time of gradient elution and the slope of the linear gradient elution curve. The advantage of the developed model is the application of an optimized two-phase training algorithm that enables the researcher to make use of the advantages of first- and second-order training algorithms in one training procedure. This results in better predictive ability, with less time required for the calculations. The number of hidden layer neurons and experimental data points used for the training set were optimized in terms of obtaining a precise and accurate retention model with respect to minimization of unnecessary experimentation and time needed for the calculation procedures. This study shows that developed, ANNs are the method of first choice for retention modelling of inorganic anions in IC.

Development of an inorganic cations retention model in ion chromatography by means of artificial neural networks with different two-phase training algorithms

This paper describes development of artificial neural network (ANN) retention model, which can be used for method development in variety of ion chromatographic applications. By using developed retention model it is possible both to improve performance characteristic of developed method and to speed up new method development by reducing unnecessary experimentation. Multilayered feed forward neural network has been used to model retention behaviour of void peak, lithium, sodium, ammonium, potassium, magnesium, calcium, strontium and barium in relation with the eluent flow rate and concentration of methasulphonic acid (MSA) in eluent. The probability of finding the global minimum and fast convergence at the same time were enhanced by applying a two-phase training procedure. The developed two-phase training procedure consists of both first and second order training. Several training algorithms were applied and compared, namely: back propagation (BP), delta-bar-delta, quick propagation, conjugate gradient, quasi Newton and Levenberg-Marquardt. It is shown that the optimized two-phase training procedure enables fast convergence and avoids problems arisen from the fact that every new weight initialization can be regarded as a new starting position and yield irreproducible neural network if only second order training is applied. Activation function, number of hidden layer neurons and number of experimental data points used for training set were optimized in order to insure good predictive ability with respect to speeding up retention modelling procedure by reducing unnecessary experimental work. The predictive ability of optimized neural networks retention model was tested by using several statistical tests. This study shows that developed artificial neural network are very accurate and fast retention modelling tool applied to model varied inherent non-linear relationship of retention behaviour with respect to mobile phase parameters.

Zeroing polynomials using modified constrained neural network approach

This paper proposes new modified constrained learning neural root finders (NRFs) of polynomial constructed by backpropagation network (BPN). The technique is based on the relationships between the roots and the coefficients of polynomial as well as between the root moments and the coefficients of the polynomial. We investigated different resulting constrained learning algorithms (CLAs) based on the variants of the error cost functions (ECFs) in the constrained BPN and derived a new modified CLA (MCLA), and found that the computational complexities of the CLA and the MCLA based on the root-moment method (RMM) are the order of polynomial, and that the MCLA is simpler than the CLA. Further, we also discussed the effects of the different parameters with the CLA and the MCLA on the NRFs. In particular, considering the coefficients of the polynomials involved in practice to possibly be perturbed by noisy sources, thus, we also evaluated and discussed the effects of noises on the two NRFs. Finally, to demonstrate the advantage of our neural approaches over the nonneural ones, a series of simulating experiments are conducted.

Feature classes for 1D, 2nd order image structure arise from natural image maximum likelihood statistics

Much is understood of how quantitative aspects of image structure are measured by VI simple cells, but less about how qualitative structure is determined from these measurements. We review Geometric Texton Theory (GTT) that aims to describe this step from quantitative to qualitative. GTT proposes that qualitative feature categories arise through consideration of the maximum likelihood (ML) explanations of image measurements. It posits that a pair of output vectors of an ensemble of co-localised neurons signal the same feature category if and only if the corresponding ML explanations are qualitatively similar. We present mathematical and empirical results relevant to GTT for the limited case of measurement by 1D filters of up to 2nd order. The mathematical results identify the simplest explanations for measurements by such filters, while the empirical results identify the ML. We find that the ML explanations are not the most simple under any of the definitions of simple that we examined. However, the ML explanations do have properties predicted by GTT. In particular they change rapidly and qualitatively for certain narrow regions of measurement space, while remaining qualitative stable between those transition regions. Three feature categories arise naturally from the data: light bars, dark bars and edges. The results are consistent with GTT.

An incremental approach to genetic-algorithms-based classification

Incremental learning has been widely addressed in the machine learning literature to cope with learning tasks where the learning environment is ever changing or training samples become available over time. However, most research work explores incremental learning with statistical algorithms or neural networks, rather than evolutionary algorithms. The work in this paper employs genetic algorithms (GAs) as basic learning algorithms for incremental learning within one or more classifier agents in a multiagent environment. Four new approaches with different initialization schemes are proposed. They keep the old solutions and use an "integration" operation to integrate them with new elements to accommodate new attributes, while biased mutation and crossover operations are adopted to further evolve a reinforced solution. The simulation results on benchmark classification data sets show that the proposed approaches can deal with the arrival of new input attributes and integrate them with the original input space. It is also shown that the proposed approaches can be successfully used for incremental learning and improve classification rates as compared to the retraining GA. Possible applications for continuous incremental training and feature selection are also discussed.

The dynamic wave expansion neural network model for robot motion planning in time-varying environments

We introduce a new type of neural network--the dynamic wave expansion neural network (DWENN)--for path generation in a dynamic environment for both mobile robots and robotic manipulators. Our model is parameter-free, computationally efficient, and its complexity does not explicitly depend on the dimensionality of the configuration space. We give a review of existing neural networks for trajectory generation in a time-varying domain, which are compared to the presented model. We demonstrate several representative simulative comparisons as well as the results of long-run comparisons in a number of randomly-generated scenes, which reveal that the proposed model yields dominantly shorter paths, especially in highly-dynamic environments.

NEUROM: a ROM based RNS digital neuron

In this work, a fast digital device is defined, which is customized to implement an artificial neuron. Its high computational speed is obtained by mapping data from floating point to integer residue representation, and by computing neuron functions through residue arithmetic operations, with the use of table look-up techniques. Specifically, the logic design of a residue neuron is described and complexity figures of area occupancy and time consumption of the proposed device are derived. The approach was applied to the logic design of a residue neuron with 12 inputs and with a Residue Number System defined in such a way as to attain an accuracy better than or equal to the accuracy of a 20-bit floating point system. The proposed design (NEUROM) exploits the RNS carry independence property to speed up computations, in addition it is very suitable for using look-up tables. The response time of our device is about 8 x T(ACC), where T(ACC) is the ROM access time. With a value of T(ACC) close to the 10 ns allowed by the current ROM technology, the proposed neuron responds within 80 ns, NEUROM is therefore the neuron device proposed in the literature which allows for maximum throughput. Moreover, when a pipeline mode of operation is adopted, the pipeline delay can assume a value as low as about 14 ns. In the case study considered, the total amount of ROM is about 5.55 Mbits. Thus, using current technology, it is possible to integrate several residue neurons into a single VLSI chip, thereby enhancing chip throughput. The paper also discusses how this amount of memory could be reduced, at the expense of the response time.

An Evaluation of Contrast Enhancement Techniques for Mammographic Breast Masses

The main aim of this paper is to propose a novel set of metrics that measure the quality of the image enhancement of mammographic images in a computer-aided detection framework aimed at automatically finding masses using machine learning techniques. Our methodology includes a novel mechanism for the combination of the metrics proposed into a single quantitative measure. We have evaluated our methodology on 200 images from the publicly available digital database for screening mammograms. We show that the quantitative measures help us select the best suited image enhancement on a per mammogram basis, which improves the quality of subsequent image segmentation much better than using the same enhancement method for all mammograms.

Understanding the role of communication and conflict on reengineering team development

Purpose

Aims to investigate the importance of communication and conflict on the development of reengineering teams and attempts to answer the following research questions: is the team development of a reengineering team affected by conflict and communication, and if so, what are the types of communication and conflict that do affect it?

Design/methodology/approach

Presents a theoretical perspective with a literature review of such aspects as conflict and communication. Details research methodology and then utilises evidence from a single in‐depth case study. Explains how the analysis occured and discusses the findings.

Findings

It was found that conflicts are not productive for the organisation and therefore conflict resolution is sought. In order to resolve the conflicts existent within the teams, the role of face‐to‐face communication was considered to be pertinent.

Originality/value

For industry that is always searching for ways to curtail excessive costs, an understanding of the issues of conflict, team development and communication is offered. This means that by examining the guidelines offered within this paper, a convenient method of identifying and solving these issues is provided.

A knowledge-based framework for image enhancement in aviation security

The main aim of this paper is to present a knowledge-based framework for automatically selecting the best image enhancement algorithm from several available on a per image basis in the context of X-ray images of airport luggage. The approach detailed involves a system that learns to map image features that represent its viewability to one or more chosen enhancement algorithms. Viewability measures have been developed to provide an automatic check on the quality of the enhanced image, i.e., is it really enhanced? The choice is based on ground-truth information generated by human X-ray screening experts. Such a system, for a new image, predicts the best-suited enhancement algorithm. Our research details the various characteristics of the knowledge-based system and shows extensive results on real images.

Phase synchronization for the recognition of mental tasks in a brain-computer interface

Brain-computer interfaces (BCIs) may be a future communication channel for motor-disabled people. In surface electroencephalogram (EEG)-based BCIs, the extracted features are often derived from spectral estimates and autoregressive models. We examined the usefulness of synchronization between EEG signals for classifying mental tasks. To this end, we investigated the performance of features derived from the phase locking value (PLV) and from the spectral coherence and compared them to the classification rates resulting from the power densities in alpha, beta1, beta2, and 8-30-Hz frequency bands. Five recordings of 60 min, acquired from three subjects while performing three different mental tasks, were analyzed offline. No artifacts were removed or rejected. We noticed significant differences between PLV and mean spectral coherence. For sole use of synchronization measures, classification accuracies up to 62% were achieved. In general, the best result was obtained combining phase synchronization measures with alpha power spectral density estimates. The results demonstrate that phase synchronization provides relevant information for the classification of spontaneous EEG during mental tasks.

A constructive approach for finding arbitrary roots of polynomials by neural networks

This paper proposes a constructive approach for finding arbitrary (real or complex) roots of arbitrary (real or complex) polynomials by multilayer perceptron network (MLPN) using constrained learning algorithm (CLA), which encodes the a priori information of constraint relations between root moments and coefficients of a polynomial into the usual BP algorithm (BPA). Moreover, the root moment method (RMM) is also simplified into a recursive version so that the computational complexity can be further decreased, which leads the roots of those higher order polynomials to be readily found. In addition, an adaptive learning parameter with the CLA is also proposed in this paper; an initial weight selection method is also given. Finally, several experimental results show that our proposed neural connectionism approaches, with respect to the nonneural ones, are more efficient and feasible in finding the arbitrary roots of arbitrary polynomials.

Possible Conflicts: A Compilation Technique for Consistency-Based Diagnosis

Consistency-based diagnosis is one of the most widely used approaches to model-based diagnosis within the artificial intelligence community. It is usually carried out through an iterative cycle of behavior prediction, conflict detection, candidate generation, and candidate refinement. In that process conflict detection has proven to be a nontrivial step from the theoretical point of view. For this reason, many approaches to consistency-based diagnosis have relied upon some kind of dependency-recording. These techniques have had different problems, specially when they were applied to diagnose dynamic systems. Recently, offline dependency compilation has established itself as a suitable alternative approach to online dependency-recording. In this paper we propose the possible conflict concept as a compilation technique for consistency-based diagnosis. Each possible conflict represents a subsystem within system description containing minimal analytical redundancy and being capable to become a conflict. Moreover, the whole set of possible conflicts can be computed offline with no model evaluation. Once we have formalized the possible conflict concept, we explain how possible conflicts can be used in the consistency-based diagnosis framework, and how this concept can be easily extended to diagnose dynamic systems. Finally, we analyze its relation to conflicts in the general diagnosis engine (GDE) framework and compare possible conflicts with other compilation techniques, especially with analytical redundancy relations (ARRs) obtained through structural analysis. Based on results from these comparisons we provide additional insights in the work carried out within the BRIDGE community to provide a common framework for model-based diagnosis for both artificial intelligence and control engineering approaches.

The loss of self-boundaries: towards a neuromolecular theory of schizophrenia

I start out with the hypothesis that the basic symptoms of schizophrenia are caused by a loss of self-boundaries. Phenomenologically, schizophrenic symptoms are based on the inability of the brain to delimit conceptual boundaries. At the cellular level in the brain, I have in previous work attributed a spatio-temporal boundary setting function to the glial cells such that glial cells determine the grouping of neurons into functional units. Mutations in genes that result in non-splicing of introns can produce aberrant versions of neurotransmitter receptors that lack protein domains encoded by entire exons and can also have protein sequence encoded by introns that have not been properly spliced out. I propose that such "chimeric" receptors are generated in glial cells and that they cannot interact properly with their cognate neurotransmitters. The glia will then lose their inhibitory function with respect to the information processing within neuronal networks. The loss of glial boundary-setting may result in a 'borderless' generalization of information processing such that the structuring of the brain in functional domains is almost completely lost. This loss of glial boundary setting could be an explanation of the loss of self-boundaries in schizophrenia.

Natural image profiles are most likely to be step edges

We introduce Geometric Texton Theory (GTT), a theory of categorical visual feature classification that arises through consideration of the metamerism that affects families of co-localised linear receptive-field operators. A refinement of GTT that uses maximum likelihood (ML) to resolve this metamerism is presented. We describe a method for discovering the ML element of a metamery class by analysing a database of natural images. We apply the method to the simplest case--the ML element of a canonical metamery class defined by co-registering the location and orientation of profiles from images, and affinely scaling their intensities so that they have identical responses to 1-D, zeroth- and first-order, derivative of Gaussian operators. We find that a step edge is the ML profile. This result is consistent with our proposed theory of feature classification.

Massively Parallel Classification of Single-Trial EEG Signals Using a Min-Max Modular Neural Network

This paper presents a method for classifying single-trial electroencephalogram (EEG) signals using min-max modular neural networks implemented in a massively parallel way. The method has three main steps. First, a large-scale, complex EEG classification problem is simply divided into a reasonable number of two-class subproblems, as small as needed. Second, the two-class subproblems are simply learned by individual smaller network modules in parallel. Finally, all the individual trained network modules are integrated into a hierarchical, parallel, and modular classifier according to two module combination laws. To demonstrate the effectiveness of the method, we perform simulations on fifteen different four-class EEG classification tasks, each of which consists of 1491 training and 636 test data. These EEG classification tasks were created using a set of non-averaged, single-trial hippocampal EEG signals recorded from rats; the features of the EEG signals are extracted using wavelet transform techniques. The experimental results indicate that the proposed method has several attractive features. 1) The method is appreciably faster than the existing approach that is based on conventional multilayer perceptrons. 2) Complete learning of complex EEG classification problems can be easily realized, and better generalization performance can be achieved. 3) The method scales up to large-scale, complex EEG classification problems.

Class Decomposition for GA-Based Classifier Agents—A Pitt Approach

This paper proposes a class decomposition approach to improve the performance of GA-based classifier agents. This approach partitions a classification problem into several class modules in the output domain, and each module is responsible for solving a fraction of the original problem. These modules are trained in parallel and independently, and results obtained from them are integrated to form the final solution by resolving conflicts. Benchmark classification data sets are used to evaluate the proposed approaches. The experiment results show that class decomposition can help achieve higher classification rate with training time reduced.

Consolidation of Long-Term Memory: Evidence and Alternatives

Memory loss in retrograde amnesia has long been held to be larger for recent periods than for remote periods, a pattern usually referred to as the Ribot gradient. One explanation for this gradient is consolidation of long-term memories. Several computational models of such a process have shown how consolidation can explain characteristics of amnesia, but they have not elucidated how consolidation must be envisaged. Here findings are reviewed that shed light on how consolidation may be implemented in the brain. Moreover, consolidation is contrasted with alternative theories of the Ribot gradient. Consolidation theory, multiple trace theory, and semantization can all handle some findings well but not others. Conclusive evidence for or against consolidation thus remains to be found.

The self-composing brain: towards a glial-neuronal brain theory

A brain model is proposed which describes its structural organization and the related functions as compartments organized in time and space. On a molecular level the negative feedback loops of clock-controlled genes are interpreted as compartments. This spatio-temporal operational principle may also work on the cellular level as glial-neuronal interactions, wherein glia have a spatio-temporal boundary setting function. The synchronization of the multi-compartmental operations of the brain is compared to the harmonization in a symphony and appears as an integrated behavior of the whole organism, defined as modes of behavior. For explanation of the principle of harmonization, an example from Schubert's Symphony No. 8 has been chosen. While harmonization refers to the synchronization of diverse systems, it seems appropriate to select the brain of a composer and the structure of musical composition as a paradigm towards a glial-neuronal brain theory. Finally, some limitations of experimental brain research are discussed and robotics are proposed as a promising alternative.

Transformations of sigma-pi nets: obtaining reflected functions by reflecting weight matrices

This paper presents a methodology that reflected functions by reflecting the weight matrices of an artificial neural network. One of the major problems with the connectionist approach is that trained neural networks can only associate fixed sets of input-output mappings. We provide a methodology which allows the post-trained net to associate different input-output mappings. The different mappings are reflected in a horizontal axis, reflected in a vertical axis and scaling of the initial mapping. The methodology does not train the net on the different mappings but it transforms the weight matrix of the neural network. This paper describes a novel way of utilising sigma-pi neural networks. Our new methodology manipulates sigma-pi unit's weight matrices which transform the unit's output. The weights are cast in a matrix formulation, and then transformations can be performed on the weight matrix of the sigma-pi net. To test the new methodology, the following three steps were carried out on a neural network: (1) the network was trained to perform a mapping function, f; (2) the weights of the network were transformed; and (3) the network was tested to evaluate whether it performs the reflection in the vertical axis,f(ref-vert)(x) = a - f(x). This reflects the function in one dimension. A reflection transformation was used to manipulate the network's weight matrices to obtain a reflection in the vertical axis. Note that the network was not trained to perform the reflection in the vertical axis. The transformation of the weight matrix transformed the function the output performs. This article explains the theory which enables us to perform transformations of sigma-pi networks and obtain reflections of the output by reflecting the weight matrices. These transforms empower the network to perform related mapping tasks once one mapping task has been learnt. This article explains how each transformation is performed and it considers whether a set of 'standard' transformations can indeed be derived.