Rule Extraction

Rough Non-deterministic Information Analysis (RNIA) is a rough set-based data analysis framework for Nondeterministic Information Systems (NISs). RNIA-related algorithms and software tools developed so far for rule generation provide good characteristics of NISs and can be successfully applied to decision making based on non-deterministic data. In this paper, we extend RNIA by introducing stability factor that enables to evaluate rules in a more flexible way and by developing a question-answering functionality that enables decision makers to analyze data gathered in NISs in case there are no pre-extracted rules that may address specified conditions.

E-learning offers a new context for education where large amounts of information describing the continuum of the teaching-learning interactions are endlessly generated and ubiquitously available. But raw information by itself may be of no help to any of the elearning actors. The use of Data Mining methods to extract knowledge from this information can, therefore, be an adequate approach to follow in order to use the obtained knowledge to fit the educational proposal to the students' needs and requirements. In this brief study we use an extension of Fuzzy Inductive Reasoning methodology to extract comprehensible, actionable and reasonable set of rules describing the students' learning behavior. The obtained rules can be used to improve the system understanding and to provide valuable information to tutors about the course performance. The extraction rules model presented in this research is applied to a real virtual campus graduate course of the Center of Studies in Communication and Educational Technologies (CECTE, as Spanish acronym).

E-learning offers a new context for education where large amounts of information describing the continuum of the teaching-learning interactions are endlessly generated and ubiquitously available. But raw information by itself may be of no help to any of the elearning actors. The use of Data Mining methods to extract knowledge from this information can, therefore, be an adequate approach to follow in order to use the obtained knowledge to fit the educational proposal to the students' needs and requirements. In this brief study we use an extension of Fuzzy Inductive Reasoning methodology to extract comprehensible, actionable and reasonable set of rules describing the students' learning behavior. The obtained rules can be used to improve the system understanding and to provide valuable information to tutors about the course performance. The extraction rules model presented in this research is applied to a real virtual campus graduate course of the Center of Studies in Communication and Educational Technologies (CECTE, as Spanish acronym).

Genetic algorithm is one of the commonly used approaches on data mining. In this paper, we put forward a genetic algorithm approach for classification problems. Binary coding is adopted in which an individual in a population consists of a fixed number of rules that stand for a solution candidate. The evaluation function considers four important factors which are error rate, entropy measure, rule consistency and hole ratio, respectively. Adaptive asymmetric mutation is applied by the self-adaptation of mutation inversion probability from 1-0 (0-1). The generated rules are not disjoint but can overlap. The final conclusion for prediction is based on the voting of rules and the classifier gives all rules equal weight for their votes. Based on three databases, we compared our approach with several other traditional data mining techniques including decision trees, neural networks and naive bayes learning. The results show that our approach outperformed others on both the prediction accur...

Artificial neural networks is one of the most commonly used machine learning algorithms in medical applications. However, they are still not used in practice in the clinics partly due to their lack of explanatory capacity. We compare two case-based explanation methods to two trained physicians on analysis of electrocardiogram (ECG) data from patients with a suspected acute coronary syndrome (ACS). The median overlaps of the top 5 selected features between the two physicians, and a given physician and a method, were initially low. Using a correlation analysis of the features the median overlap increased to values typically in the range 2-3. In conclusion, both our case-based methods generate explanations somewhat similar to those of trained expert physicians on the problem of diagnosing ACS from ECG data.

The neural networks are successfully applied to many applications in different domains. However, due to the results made by the neural networks are difficult to explain the decision process of neural networks is supposed as a black box. The explanation of reasoning is important to some applications such like credit approval application and medical diagnosing software. Therefore, the rule extraction algorithm is becoming more and more important in explaining the extracted rules from the neural networks. In this paper, a decompositional algorithm is analyzed and designed to extract rules from neural networks. The algorithm is simple but efficient; can reduce the extracted rules but improve the efficiency of the algorithm at the same time. Moreover, the algorithm is compared to the other two algorithms, M-of-N and Garcez, by solving the MONK's problem.

This paper presents a new approach to fuzzy rulebased modeling of nonlinear systems from numerical data. The novelty of the approach lies in the way of input partitioning and in the syntax of the rules. This paper introduces interpretable relational antecedents that incorporate local linear interactions between the input variables into the inference process. This modification improves the approximation quality and allows for limiting the number of rules. Additionally, the resulting linguistic description better captures the system characteristics by exposing the interactions between the input variables.

Back-propagation learning (BP) is known for its serious limitations in generalising knowledge from certain types of learning material. BP-SOM is an extension of BP which overcomes some of these limitations. BP-SOM is a combination of a multi-layered feed-forward network (MFN) trained with BP, and Kohonen's self-organising maps (SOMs). In earlier reports, it has been shown that BP-SOM improved the generalisation performance whereas it decreased simultaneously the number of necessary hidden units without loss ...

Recent years have shown the need of an automated p r ocess to discover interesting and hidden patterns in real-world databases, handling large volumes of data. This sort of process implies a lot of computational power, memory and disk I O, which can only be p r ovided b y p arallel computers. Our work contributes with a solution that integrates a machine learning algorithm, parallelism and a tightly-coupled use of a DBMS system, addressing performance problems with parallel processing and data fragmentation.

In a text-to-speech system, a transcription of each word can be either retrieved from the dictionary, or generated by rules or some statistical means. Though the dictionary-based approach can produce the most accurate result, a letter-to-sound conversion module is ...

Through conceptual examples and demonstrations, we argue that the symbiotic combination of the Internet and humans will result in a significant enhancement of the previously existing, self-organizing social structure of humans. The combination of the unique capabilities of intelligent, distributed information systems (the relatively loss-less transmission and capturing of detailed signatures) with the unique capabilities of humans (processing and analysis of complex, but limited, systems) will enable essential problem solving within our increasingly complex world. The capability may allow solutions that are not achievable directly by individuals, organizations or governments.

Classification is one of the data mining problems receiving great attention recently in the database community. This paper presents an approach to discover symbolic classification rules using neural networks. Neural networks have not been thought suited for data mining because how the classifications were made is not explicitly stated as symbolic rules that are suitable for verification or interpretation by humans. With the proposed approach, concise symbolic rules with high accuracy can be extracted from a neural network. The network is first trained to achieve the required accuracy rate. Redundant connections of the network are then removed by a network pruning algorithm. The activation values of the hidden units in the network are analyzed, and classification rules are generated using the result of this analysis. The effectiveness of the proposed approach is clearly demonstrated by the experimental results on a set of standard data mining test problems.

Although backpropagation neural networks generally predict better than decision trees do for pattern classification problems, they are of-ten regarded as black boxes, ie, their predic-tions are not as interpretable as those of deci-sion trees. This paper argues that ...

Classi cation, which involves nding rules that partition a given data set into disjoint groups, is one class of data mining problems. Approaches proposed so far for mining classication rules for large databases are mainly decision tree based symbolic learning methods. The connectionist approach based on neural networks has been thought not well suited for data mining. One of the major reasons cited is that knowledge generated by neural networks is not explicitly represented in the form of rules suitable for veri cation or ...

Previously, we hypothesized that overfitting occurs in the brain networks of the patients with autism. In this study, we implement this idea in an artificial neural network (ANN). A set of tasks, named configural grouping tests, used by Mottron and colleagues in 2003, were picked out to be applied on a typical ANN which we designed. The visual stimuli were fed as the input; over time, the output error was monitored to evaluate the generalization capability of the network. The ANN training results of 6 letters were represented and the over-fitting was observed in all of them. In the learning process, before over-fitting (over learning) occurrence, testing error diagram decreases with decay in training error which indicates that ANN was trained well. This condition simulates the behavior of a healthy person. By continuing the learning process, testing error begins to increase and over-fitting occurs in the ANN, i.e. the generalization capability of the network decreases which simulate...

Land cover classification using multispectral satellite image is a very challenging task with numerous practical applications. We propose a multi-stage classifier that involves fuzzy rule extraction from the training data and then generation of a possibilistic label vector for each pixel using the fuzzy rule base. To exploit the spatial correlation of land cover types we propose four different information aggregation methods which use the possibilistic class label of a pixel and those of its eight spatial neighbors for making the final classification decision. Three of the aggregation methods use Dempster-Shafer theory of evidence while the remaining one is modeled after the fuzzy k-NN rule. The proposed methods are tested with two benchmark seven channel satellite images and the results are found to be quite satisfactory. They are also compared with a Markov random field (MRF) model-based contextual classification method and found to perform consistently better.

Most methods of fuzzy rule-based system identification (SI) either ignore feature analysis or do it in a separate phase. This paper proposes a novel neuro-fuzzy system that can simultaneously do feature analysis and SI in an integrated manner. It is a five-layered feed-forward network for realizing a fuzzy rule-based system. The second layer of the net is the most important one, which along with fuzzification of the input also learns a modulator function for each input feature. This enables online selection of important features by the network. The system is so designed that learning maintains the nonnegative characteristic of certainty factors of rules. The proposed network is tested on both synthetic and real data sets and the performance is found to be quite satisfactory. To get an "optimal" network architecture and to eliminate conflicting rules, nodes and links are pruned and then the structure is retrained. The pruned network retains almost the same level of performance as that of the original one.

Neural networks have been applied in various domain including science, commerce, medicine, and industry. However, The knowledge learned by a trained neural network is difficult to understand. This paper proposes a Boolean algebra based algorithm to extract comprehensible Boolean rules from supervised feed-forward neural networks to uncover the black-boxed knowledge. This algorithm is called the BAB-BB rule extraction algorithm, which stands for a Boolean algebra based rule extraction algorithm for neural networks with binary and bipolar inputs. Decomposition techniques and interval arithmetic are used in the algorithm. First, each neuron associated with its inputs is analyzed and a Boolean function, describing the activation rule from its inputs to the neuron, is derived. These Boolean functions are merged into an aggregated Boolean rule according to the network topology. The Boolean rule is then further simplified by Boolean algebra operations. During the rule extraction procedure, redundant hidden neurons can be detected and removed without affecting the original function of the neural network. Examples of unipolar and bipolar inputs are presented to demonstrate the use of our algorithm. Finally, the Exclusive OR problem is presented and solved by our algorithm. Results show that our BAB-BB algorithm is practicable and of high efficiency.

A recognized impediment to the more widespread utilization of Artificial Neural Networks (ANNs) is the absence of a capability to explain, in a humancomprehensible form, either the process by which a trained ANN arrives at a specific decisionlresult or, in general, the totality of knowledge embedded therein. Recently there has been a proliferation of techniques aimed at redressing this situation and, in particular, for extracting the knowledge embedded in trained feedforward ANNs as sets of symbolic rules. However, if the dissemination of ideas in the field of ANN rule extraction is to proceed in a systematic manner, then it is essential that U rigorous taxonomy exists for categorizing the plethora of techniques being developed. This paper shows how one of the proposed schemas for categorizing ANN rule extraction techniques is able to accommodate such recent developments in the field. In addition uttention is drawn to what are seen to be some of the key challenges in the area including the identification of factors which appear to limit what is actually achievable through the rule extraction process.

Artificial Neural Networks (ANNs) are often viewed as black box. This limits the comprehensive understanding on how it deals with input neuron/data, as well as how it reached a particular decision. Input significance analysis (ISA) refers to the process of understanding these input neurons/data. And since this work is on classification problem, hence similarly, this process can also be called feature selection; where the goal is to have a classifier that can predict accurately and at the same time, its structure is as simple as possible. This work is particularly interested with ISA methods that manipulate weights, where separately, correlations are also applied. The goal is to create feature ranking list that performed the best in the selected classifiers. For validation methods, memory recall validation and K-Fold cross-validation methods are used. The results show one classifier that uses one of the ISA methods are performing well for both validation methods.

This paper presents a technique to improve the accuracy of the predictions obtained using the Rough Set Theory (RST) in non-deterministic cases (rough cases). The RST is here applied to the data collected by the Intelligent Field Devices for identifying predictive diagnostic algorithms for machinery, plants, subsystems, or components. The data analysis starts from a historical data set recorded from the field instruments, and its final result is a set of ''if-then" rules identifying predictive maintenance functions. These functions may be used to predict if a component is going to fail or not in the next future. The prediction is obtained by applying the rules extracted with the RST algorithm on the real-time values transmitted by the field device. It may happen that some diagnoses are uncertain, in the sense that it is not possible to take a certain decision (device sound or close to fail) with a given set of data. In this paper, a new algorithm for increasing the confidence in these uncertain cases is presented. To show an example, the proposed confirmation algorithm is applied to the predictive algorithms obtained for an intelligent pressure transmitter.

The problem of relevance and the usefulness of extracted association rules is becoming of primary importance, since an overwhelming number of association rules may be derived. This paper proposes an algorithm, called GenAll, to build a formal concept lattice, in which each formal concept is "decorated" by its minimal generators. The main characteristic of this algorithm is to use a refinement process of upper cover lists to determine, in a simultaneous manner, the set of formal concepts, their underlying partial order and the set of minimal generators associated to each formal concept. Experimental results have showed that the proposed algorithm is specially efficient for dense formal contexts compared to that of Nourine et al.. Response times pointed out by GenAll algorithm largely outperform those of Nourine et al..

Artificial Neural Networks (ANNs) are able, in general and in principle, to learn complex tasks. Interpretation of models induced by ANNs, however, is often extremely difficult due to the non-linear and non-symbolic nature of the models. To enable better interpretation of the way knowledge is represented in ANNs, we present BP-SOM, a neural network architecture and learning algorithm. BP-SOM is a combination of a multi-layered feed-forward network (MFN) trained with the back-propagation learning rule (BP), and Kohonen's self-organising maps (SOMs). The involvement of the SOM in learning leads to highly structured knowledge representations both at the hidden layer and on the SOMs. We focus on a particular phenomenon within trained BP-SOM networks, viz. that the SOM part acts as an organiser of the leaqrning material into instance subsets that tend to be homogeneous with respect to both class labelling and subsets of attribute values. We show that the structured knowledge representation can either be exploited directly for rule extraction, or be used to explain a generic type of checksum solution found by the network for learning M-of-N tasks.

In France, 40 % of buildings are heated with electrical devices causing high peak load in winter. In this context, advanced control systems could improve buildings energy management. More specifically, optimal strategies have been developed using a dynamic programming method in order to shift heating load, taking advantage of the building thermal mass. However, this optimisation method is computationally intensive and can hardly be applied to real-time control. Statistical techniques can be used to derive near-optimal laws from the optimal control results. These rule extraction techniques model the relationship between explanatory variables and a response variable. This paper investigates the use of Beta regression model. This regression-based strategy was able to mimic the general characteristics of the optimisation results with a small mean bias error (-6 %) and greatly reduce computational effort (150 times faster). Given its simple mathematical formulation, it could be implemented in real-time building systems control.

In France, 40 % of buildings are heated with electrical devices causing high peak load in winter. In this context, optimal strategies (under constraints related to comfort and maximum heating power) have been developed using the dynamic programming method in order to shift electricity consumption used for heating, taking advantage of the building thermal mass. However, this exact optimisation method is computationally intensive and can hardly be applied to real-time control. Complementary statistical techniques exist that allow for the extraction of logistic decision models from the optimal control simulation results. These rule extraction techniques model the relationship between explanatory variables and a response variable. In this study, a generalised linear model was used because it is able to mimic the general characteristics of the dynamic programming results with good precision and greatly reduced computational effort (150 times faster than the dynamic programming method).

Promoter sequences are well known to play a central role in gene expression. Their recognition and assignment in silico has not consolidated into a general bioinformatics method yet. Most previously available algorithms employ and are limited to s70-dependent promoter sequences. This paper presents a new tool named BacPP, designed to recognize and predict Escherichia coli promoter sequences from background with specific accuracy for each s factor (respectively, s24, 86.9%; s28, 92.8%; s32, 91.5%; s38, 89.3%, s54, 97.0%; and s70, 83.6%). BacPP is hence outstanding in recognition and assignment of sequences according to s factor and provide circumstantial information about upstream gene sequences. This bioinformatic tool was developed by weighing rules extracted from neural networks trained with promoter sequences known to respond to a specific s factor. Furthermore, when challenged with promoter sequences belonging to other enterobacteria BacPP maintained 76% accuracy overall.

Les études concernant les méthodes d'extraction de règle de classification proposent de classer une instance, le plus souvent en se basant sur une seule règle, sans prendre en considération les règles du même type ayant un label de classe différent (par ex. : X → a, X → b). Dans certains domaines, comme par exemple le domaine médical, il est important de connaître toutes les conséquences induites par une règle donnée. Dans cette étude, nous proposons une nouvelle méthode d'extractions de règle de classification multi-labels. Elle se caractérise par sa rapidité, ceci est dû à la procédure utilisée pour l'extraction des motifs fréquents par le biais des ensembles flous. Elle se caractérise également par l'extraction des règles en fonction de leur label, les règles sont extraites à partir des ensembles de données appartenant à une même classe et non à partir de l'ensemble global des données. Cette approche permet d'extraire des règles spécifiques à chaque classe et d'extraire de nouvelles catégories de règle d'exception. La méthode développée présente une grande précision et ses performances sont comparables à celles d'autres méthodes de classification de références.

This paper presents a technique to improve the accuracy of the predictions obtained using the Rough Set Theory (RST) in non-deterministic cases (rough cases). The RST is here applied to the data collected by the Intelligent Field Devices for identifying predictive diagnostic algorithms for machinery, plants, subsystems, or components. The data analysis starts from a historical data set recorded from the field instruments, and its final result is a set of ''if-then" rules identifying predictive maintenance functions. These functions may be used to predict if a component is going to fail or not in the next future. The prediction is obtained by applying the rules extracted with the RST algorithm on the real-time values transmitted by the field device. It may happen that some diagnoses are uncertain, in the sense that it is not possible to take a certain decision (device sound or close to fail) with a given set of data. In this paper, a new algorithm for increasing the confidence in these uncertain cases is presented. To show an example, the proposed confirmation algorithm is applied to the predictive algorithms obtained for an intelligent pressure transmitter.

In the paper, the problem of extraction of complex decision rules in simple decision systems over ontological graphs is considered. The extracted rules are consistent with the dominance principle similar to that applied in the dominancebased rough set approach (DRSA). In our study, we propose to use a heuristic algorithm, utilizing the ant-based clustering approach, searching the semantic spaces of concepts presented by means of ontological graphs. Concepts included in the semantic spaces are values of attributes describing objects in simple decision systems.

A novel ant colony algorithm, mass recruitment and group recruitment based continuous ant colony optimization (MG-CACO), is proposed to solve continuous optimization problems. MG-CACO, which can capture the interdependencies between attributes and does not need discretization as a preprocessing step for optimization, was applied to extract classification rules from samples. To improve the predictive performance of the classifier, the ensemble strategy was adopted and the MG-CACO based ensemble classifier system called MG-CACO-ECS was built. Several datasets, obtained from UCI (University of California, Irvine) machine learning repository, were employed to illustrate the validity of MG-CACO-ECS. The results indicated that MG-CACO-ECS has satisfactory prediction accuracy. Furthermore, the problem of the producing area discrimination of olive oil was studied, and the obtained results demonstrated that MG-CACO-ECS has better prediction accuracy than the reported results.

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

The wave of innovation unleashed by the first user-friendly PCs in the 1980's and the Web in the 1990's seems to have gotten drowned in complexity and confusion. Software developers are scrambling to keep their systems up-to-date with all the new standards, plugins and extensions. While we constantly hear announcements of the most spectacular innovations, few seem to reach maturity. The problem is that developers tend to underestimate the complexity of the task environment: today's information systems depend on so many modules, sources of data, network connections, input and output devices that it has become impossible to predict or control their interactions. The result is software full of bugs, corrupted data, security holes, viruses, and other potentially catastrophic side-effects. Moreover, systems have become so complex that the human mind simply no longer can learn or remember all procedures needed to use them. Add to this constant change in hardware, software, protocols, data, and user expectations, mix well, and you have a recipe for chaos. The complexity bottleneck not only creates stress and confusion, it severely limits the speed of further progress. The number of possible interactions grows exponentially with the number of components. Since components and their capacity increase exponentially, overall complexity increases super-exponentially. The cognitive capacity of developers obviously increases at a much slower rate, thus lagging further and further behind. This means that large projects, such as the Semantic Web, either will get endlessly delayed, or end up with unworkable products. We need a radically different approach to overcome this bottleneck. One step forward is IBM's "autonomic computing" () initiative. IBM researchers envision systems that function largely independently from their human supervisors, adapting, correcting and repairing themselves whenever a problem occurs. IBM's uses the metaphor of the autonomic nervous system, which runs our body for us without need for conscious intervention. How they hope to achieve this is less clear, though. Their suggestions seem to center around models of feedback, adaptation and control first proposed in the 1950's. While we applaud this "cybernetic" approach to computing, we believe that an even more radical vision is needed: selforganization.

Cet article presente une application de la classification multi-labels pour la prediction des diagnostics secondaires a partir d’un diagnostic primaire connu (et d’informations complementaires) dans les donnees du PMSI(Programme de Medicalisation des Systemes d'Information). Nous commencons, dans cet article, par exposer le contexte de ces travaux et justifier l’application d’une classification multi-labels pour repondre a ce besoin. Une etude bibliographique sur la classification multi-labels est ensuite presentee en mettant l’accent sur les caracteristiques des donnees multi-labels que nous manipulons et les differentes metriques utilisees dans ce domaine pour evaluer la qualite des jeux de donnees. Nous menons ensuite une etude comparative des algorithmes de la litterature appliques au contexte de la prediction de diagnostics afin d'identifier la meilleure strategie a appliquer. Cette etude est composee de trois etapes: (i) caracterisation de la qualite des donnees multi-...

Neurocognitive approach to higher cognitive functions that bridges the gap between psychological and neural level of description is introduced. Relevant facts about the brain, working memory and representation of symbols in the brain are summarized. Putative brain processes responsible for problem solving, intuition, skill learning and automatization are described. The role of non-dominant brain hemisphere in solving problems requiring insight is conjectured. Two factors seem to be essential for creativity: imagination constrained by experience, and filtering that selects most interesting solutions. Experiments with paired words association are analyzed in details and evidence for stochastic resonance effects is found. Brain activity in the process of invention of novel words is proposed as the simplest way to understand creativity using experimental and computational means. Perspectives on computational models of creativity are discussed.

Recently there has been a lot of interest in the extraction of symbolic rules from neural networks. The work described in this paper is concerned with an evaluation and comparison of the accuracy and complexity of symbolic rules extracted from radial basis function networks and multi-layer perceptrons. Here we examine the ability of rule extraction algorithms to extract meaningful rules that describe the overall performance of a particular network. In addition, the research also highlights the suitability of a speci c neural network architecture for particular classi cation problems. The research carried out on the extracted rule quality and complexity also has a direct bearing on the use of rule extraction algorithms for data mining and knowledge discovery.

Bootstrapping techniques have significant potential for the efficient generation of linguistic resources such as electronic pronunciation dictionaries. We describe a system and an approach to bootstrapping for the development of such dictionaries, and report on experiments conducted to investigate the efficiency and effectiveness of the system, focusing on the human factors that influence the process. Encouraging results were obtained: even developers with limited linguistic experience can develop accurate pronunciation models in substantially less time than a trained pronunciation expert takes using conventional methods.

The gene expression control is a fundamental process in cellular activities, performed through the interaction of multiple regulatory mechanisms. The proper regulation of transcription is crucial for a single-cell prokaryote since its environment can change dramatically and instantly. The promoters are recognized as one of the transcription regulatory regions, since recruit the transcriptional machinery through the binding of regulatory proteins in their DNA sequences. The characterizing promoter regions in silico has difficulties, since these elements are short and degenerated, providing a high probability of finding similar sequences in other parts of the genome. Therefore, the embedding of structural characteristics can increase the accuracy of prediction methods [1-2]. In bacteria, RNApolymerase holoenzyme is responsible for promoter recognition and the gene expression starts. This enzyme consists of five subunits (2α, β, β', ω) and an additional sigma (σ) subunit factor. A collection of different σ subunits act as key regulators of bacterial gene expression. The substitution of one σ factor by another can initiate the transcription of different groups of genes [3]. A promoter sequence is characterized by the presence of two conserved DNA elements called-10 and-35 (upstream). These elements are defined according to the distance which have in relation to the transcriptional start site (position 1) and are represented by TATAAT-TTGACA nucleotides [4]. The upstream region (promoter) has distinct sequence properties compared to downstream region (non-promoter), such as differences in the structural characteristics of flexibility, stability and curvature [5]. Artificial neural networks (ANNs) have been widely used in nucleic acid sequences analysis, since they present ability to recognize and classify quantitative and qualitative patterns in data analysis [6]. This work aims to predict, recognize and characterize promoter regions recognized by sigma factor 28 (σ28) employing an approach of artificial neural networks using as input parameter curvature and flexibility data of the sequence.

The gene expression control is a fundamental process in cellular activities, performed through the interaction of multiple regulatory mechanisms. The proper regulation of transcription is crucial for a single-cell prokaryote since its environment can change dramatically and instantly. The promoters are recognized as one of the transcription regulatory regions, since recruit the transcriptional machinery through the binding of regulatory proteins in their DNA sequences. The characterizing promoter regions in silico has difficulties, since these elements are short and degenerated, providing a high probability of finding similar sequences in other parts of the genome. Therefore, the embedding of structural characteristics can increase the accuracy of prediction methods [1-2]. In bacteria, RNApolymerase holoenzyme is responsible for promoter recognition and the gene expression starts. This enzyme consists of five subunits (2α, β, β', ω) and an additional sigma (σ) subunit factor. A collection of different σ subunits act as key regulators of bacterial gene expression. The substitution of one σ factor by another can initiate the transcription of different groups of genes [3]. A promoter sequence is characterized by the presence of two conserved DNA elements called-10 and-35 (upstream). These elements are defined according to the distance which have in relation to the transcriptional start site (position 1) and are represented by TATAAT-TTGACA nucleotides [4]. The upstream region (promoter) has distinct sequence properties compared to downstream region (non-promoter), such as differences in the structural characteristics of flexibility, stability and curvature [5]. Artificial neural networks (ANNs) have been widely used in nucleic acid sequences analysis, since they present ability to recognize and classify quantitative and qualitative patterns in data analysis [6]. This work aims to predict, recognize and characterize promoter regions recognized by sigma factor 28 (σ28) employing an approach of artificial neural networks using as input parameter curvature and flexibility data of the sequence.

Due to their capability of dealing with nonlinear In the last years, Formal Concept Analysis (FCA) has been pro- problems, Artificial Neural Networks (ANN) are widely used with posed as a powerful tool in knowledge representation and ex- several purposes. Once trained, they are also capable of solving traction. In this work, an alternative method, based on FCA unprecedented situations, keeping tolerable errors in their out-. . ' . . o puts. However, ANN are considered essentially "black boxes". ospropoed wh t Therefore, humans can not assimilate the knowledge kept by those firom trained ANN. nets, since such knowledge is implicitly represented by their con-This paper is organized in four sections. In the second one, fun- nection weights. In this paper, a new approach to extract knowl-damental theories used in this work are presented. In the third edge rules from ANN previously trained through Formal Concept section, rule extraction by means of FCA is presented. Finally, Analysis is presented. The method allows to the knowledge en-. ' . . gineer understand the industrial process that is being analyzed, n the last section, conclusions are presented through implications rules of the type if ... then. As an exam- ple of application a solar energy system is considered. The rules II. FUNDAMENTALS obtained are validated through an expert domain.

Neurocognitive approach to higher cognitive functions that bridges the gap between psychological and neural level of description is introduced. Relevant facts about the brain, working memory and representation of symbols in the brain are summarized. Putative brain processes responsible for problem solving, intuition, skill learning and automatization are described. The role of non-dominant brain hemisphere in solving problems requiring insight is conjectured. Two factors seem to be essential for creativity: imagination constrained by experience, and filtering that selects most interesting solutions. Experiments with paired words association are analyzed in details and evidence for stochastic resonance effects is found. Brain activity in the process of invention of novel words is proposed as the simplest way to understand creativity using experimental and computational means. Perspectives on computational models of creativity are discussed.

The present disclosure is directed to a novel system for performing online reconfiguration of a neural network. Once a neural network has been implemented into a production environment, the system may use underlying construction logic to perform an in-situ reconfiguration of neural network elements while the neural network is live. The system may accomplish the reconfiguration by modifying the architecture of the neural network and/or performing adversarial training and/or retraining. In this way, the system may provide a way increase the performance of the neural network over time along one or more performance parameters or metrics.

The present patent is directed to a novel system for a self-constructing deep neural network. The system may comprise a hybrid logic library which contains the building structures needed to construct the neural network, which may include both traditional logic and memory structures as well as learning structures. In constructing the neural network from library structures, the system may use an algorithm to iteratively improve the performance of the neural network. In this way, the system may provide a way to generate complex neural networks that become increasingly optimized over time.

Breakthroughs in information and communication technology establish payment-collection technologies like Debit and Credit card systems at the level where their rapid penetration in commercial market have led to an everlarger share of the world payment system and leading to a higher rate of stolen account numbers and subsequent losses by banks. Improved fraud detection thus has become essential to maintain the viability of the payment system, especially for e-commerce. This tremendous growth in databases, fraud detection and payment security systems has spawned a pressing need for new techniques and tools that can intelligently and automatically transform data into useful information and knowledge. In recent years, data-mining (DM) has become one of the most valuable tools for extracting and manipulating data and for establishing patterns in order to produce useful information for decision-making. This paper introduces three important data mining techniques J48, Naive Bayes and One-R...

The paper focuses on the problem of rule extraction from neural networks, with the aim of transforming the knowledge captured in a trained neural network into a familiar form for human user. The ultimate purpose for us is to develop human friendly shells for neural network based systems. In the first part of the paper it is presented an approach on extracting traditional crisp rules out of the neural networks, while the last part of the paper presents how to transform the neural network into a set of fuzzy rules using an interactive fuzzy operator. The rules are extracted from ordinary neural networks, which have not a structure that facilitate the rule extraction. The neural network trained with the well known Iris data set was considered as benchmark problem.

The overall purpose of this study is to develop a prototype radiologic consultation system. The system should provide a second diagnostic opinion based on similar cases, incorporating the experience of many radiologists, their diagnostic rules and a data base of previous cases. The system allows a radiologist to enter the description of a particular case using the lexicon such as

The overall purpose of this study is to develop a prototype radiological consultation system. We concentrate our work on prototype software environment for the system. The system provides a second diagnostic opinion based on similar cases, incorporating the experience of radiologists, their diagnostic rules and a database of previous cases. The system allows a radiologist to enter the description of a particular case using the lexicon such as BI-RADS of American College of Radiology and retrieve the second diagnostic opinion (probable diagnosis) for a given case. The system also allows a radiologist to get other important information too. These advances are based on a new computational intelligence technique and first-order logic [Mitchell, 1997]. We implemented a rule-based prototype diagnostic system. Two experimental Internet versions are currently available on the web and are under testing and evaluation of design. The diagnosis is based on the opinions of radiologists in combination with the statistically significant diagnostic rules extracted from the 2 available database.

Uses a parametric statistical framework to understand the effect of input representation on performance for nonlinear prediction of time series. In particular, considerations of input representation lead directly to choices between feedforward and recurrent neural networks. It is shown that feedforward networks are nonlinear autoregressive models and that recurrent networks can model a larger class of processes, including nonlinear autoregressive moving average models. For some processes, feedback allows recurrent networks to achieve better predictions than can be made with a feedforward network with a finite number of inputs. The results are confirmed on a problem in power system regional load forecasting

Presence of duplicate documents in the World Wide Web adversely affects crawling, indexing and relevance, which are the core building blocks of web search. In this paper, we present a set of techniques to mine rules from URLs and utilize these learnt rules for de-duplication using just URL strings without fetching the content explicitly. Our technique is composed of mining the crawl logs and utilizing clusters of similar pages to extract specific rules from URLs belonging to each cluster. Preserving each mined rules for de-duplication is not efficient due to the large number of specific rules. We present a machine learning technique to generalize the set of rules, which reduces the resource footprint to be usable at web-scale. The rule extraction techniques are robust against web-site specific URL conventions. We demonstrate the effectiveness of our techniques through experimental evaluation.

Neurocognitive approach to higher cognitive functions that bridges the gap between psychological and neural level of description is introduced. Relevant facts about the brain, working memory and representation of symbols in the brain are summarized. Putative brain processes responsible for problem solving, intuition, skill learning and automatization are described. The role of non-dominant brain hemisphere in solving problems requiring insight is conjectured. Two factors seem to be essential for creativity: imagination constrained by experience, and filtering that selects most interesting solutions. Experiments with paired words association are analyzed in details and evidence for stochastic resonance effects is found. Brain activity in the process of invention of novel words is proposed as the simplest way to understand creativity using experimental and computational means. Perspectives on computational models of creativity are discussed.