Gene Regulatory Networks (GRNs) represent the causal relations among the genes and provide insight on the cellular functions and the mechanism of the diseases. GRNs can be inferred from gene expression data by a number of algorithms, e.g. Boolean networks, Bayesian networks, and differential equations. While reliable inference of GRNs is still an open problem, new algorithms need to be developed. Fuzzy Cognitive Maps (FCMs) is used to represent GRNs in this paper. Most of the FCM learning algorithms are able to learn FCMs with less than 40 nodes. A new algorithm that is able to learn FCMs with more than 100 nodes is proposed. The proposed method is based on Ant Colony Optimization (ACO). A decomposed approach is proposed to reduce the dimension of the problem; therefore the FCM learning algorithm is more scalable (the dimension of the problem to be solved in one ACO run equals to the number of nodes or genes). The proposed approach is tested on data from DREAM project. The experiment results suggest the proposed approach outperforms several other algorithms.

Fuzzy Cognitive Maps (FCMs) are a flexible modeling technique with the goal of modeling causal relationships. Traditionally FCMs are developed by experts. We need to learn FCMs directly from data when expert knowledge is not available. The FCM learning problem can be described as the minimization of the difference between the desired response of the system and the estimated response of the learned FCM model. Learning FCMs from data can be a difficult task because of the large number of candidate FCMs. A FCM learning algorithm based on Ant Colony Optimization (ACO) is presented in order to learn FCM models from multiple observed response sequences. Experiments on simulated data suggest that the proposed ACO based FCM learning algorithm is capable of learning FCM with at least 40 nodes. The performance of the algorithm was tested on both single response sequence and multiple response sequences. The test results are compared to several algorithms, such as genetic algorithms and nonlinear Hebbian learning rule based algorithms. The performance of the ACO algorithm is better than these algorithms in several different experiment scenarios in terms of model errors, sensitivities and specificities. The effect of number of response sequences and number of nodes is discussed.

Studies in the health sciences often seek to discover cause-effect relationships among observed variables of interest, for example: treatments, exposures, preconditions, and outcomes. Consequently, causal modeling and causal discovery are central to medical science. In order to algorithmically consider causal relations, the relations must be placed into a representation that supports manipulation and discovery. Knowledge of at least some causal effects is inherently imprecise or approximate. The most widespread causal representation is directed acyclic graphs (DAGs). However, DAGs are limited in what they can represent. Another graph methodology, fuzzy cognitive maps (FCMs) hold promise as a model that overcomes some of the difficulties found in other approaches. This paper considers causality and suggests fuzzy cognitive maps as a useful causal representation methodology.

An important question is how to determine if a person is exhibiting suicidal tendencies in behavior, speech, or writing. This paper demonstrates a method of analyzing written material to determine whether or not a person is suicidal. The method involves an analysis of word frequencies that are used in a fuzzy cognitive map. The fuzzy cognitive map determines if there are suicidal tendencies. The method could have substantial potential in suicide prevention as well as in other forms of sociological behavior studies that also might exhibit their own identifying patterns.

The target of many studies in the health sciences is the discovery of cause-effect relationships among observed variables of interest, for example: treatments, exposures, preconditions, and outcomes. Causal modeling and causal discovery are central to medical science. In order to algorithmically consider causal relations, the relations must be placed into a representation that supports manipulation. Knowledge of at least some causal effects is imprecise. The most widespread causal representation is directed acyclic graphs (DAGs). However, DAGs are severely limited in what portion of the common sense world they can represent. Another network methodology, fuzzy cognitive maps hold promise. This paper considers the needs of commonsense causality and suggests Fuzzy Cognitive Maps as a useful methodology.

Ontological structures play a major role in the semantic Web; they became the target of an extensive research over the last decade. An important advance was the attempt to employ fuzzy sets and fuzzy logic techniques in representing ontologies for intrinsically vague domains of interest where most of human knowledge cannot be expressed in crisp logical formulas. Although the fuzzy approach alleviates the crispness problem, it does not deal satisfactory with the ambiguity caused by deficient discernibility of objects. We consider building rough approximations of fuzzy concepts and relations to measure the roughness of generated ontologies. The key objective of this work is to roughly approximate fuzzy concepts, entailments and sub-sumptions of a given ontology to have a better understanding of both ontology's quality and boundaries of usage. This will be useful for estimating appropriateness of existing ontologies for reasoning in incomplete domains.

This paper explores delegation decisions predicated on models of trust and autonomy among agents. In socially rich environments, trust and autonomy of artificial agents are key attributes for rational delegation decisions. Social agents are affected

Recognizing and developing causal relationships is essential for reasoning; it forms the basis for acting intelligently in the world. Causal knowledge provides a deep understanding of a system; and, the potential control over a system that comes from being able to predict action's consequences. Knowledge of at least some relationships is inherently imprecise. Causal complexes are groupings of smaller causal relations that can make up a larger-grained causal object. Usually, commonsense reasoning is more successful in reasoning about a few large-grained events than many finer-grained events. However, larger-grained causal objects are necessarily more imprecise. Often, a network represents a causal relationship with conditioned edges (probability, possibility, randomness, etc.). Various representational graphs and models can be used. Needed necessary descriptions include cycles, including mutual dependencies, both with non-cumulative effects and cumulative effects. Without cyclic descriptions, there would be an incomplete representation of the variety and wealth of causal constructions used in science as well as in everyday life. Directed Bayesian causal networks have received significant attention; they have a significant weakness in that they do not allow cycles; they have other significant restrictions, including Markoff independence conditions. This paper discusses: causality, complexes, granularity, imprecision, general and Bayesian causal models; the perspective reflects data mining goals.

An overview of a model for computational recognition of short jokes that are based on phonological similarity of words is presented. The jokes are taken from joke books. The model takes into account orthographic, phonological and semantic representation of words. A motivation for using this model, as well as steps for a joke recognizer, are discussed based on an example. The joke recognition is based on knowledge that is provided by an ontology. The ontology is created using entries from a children's dictionary. Words in the dictionary are instances of concepts in the ontology. Semantic relationships between concepts are added from a collection of children's texts and definitions in a children's dictionary. Any short text is considered a joke if it contains two scripts that both overlap and oppose; and if there is a pair of similar sounding words, in which the first word is an instance of a concept of the first script, and the second word is an instance of a concept of the second script.

The semantic Web is the next step in the Internet's evolution. The existing Web contains a considerable amount of data; most is weakly structured. Broadly accessing the data is difficult. Previously unseen data cannot be easily autonomously understood with out a consistent semantic framework. Viewing and organizing data using shared ontologies is essential for both sharing data and for Web site interoperability. Full autonomous or semi-autonomous discovery and development of ontologies is the only feasible way to transition the existing Web to the semantic Web. One strategy is to merge existing, vetted ontologies. The pre-merger ontologies would likely be similar in some respects and different in others. When comparing ontologies, a necessarily imprecise, approximate similarity metric will be necessary. Possibly, soft computing will provide useful tools

We are interested in autonomous or unsupervised data mining. In the broad context of our work, we were interested in partitioning data in order to increase the information within each partition. An important aspect of this was data granulation. Clustering is an important aspect of data mining; it enables us to granulize the data. However, classic clustering methods require either the number of clusters and/or the approximate cluster centers. To avoid providing guidance, we used a variation of the Mountain Method to develop clustering parameters. This gave us satisfactory cluster centers. The estimated clusters' centers were then used in the data mining algorithm to develop and refine partitions. Experimental results of using the method on some data sets are reported. © 2005 Wiley Periodicals, Inc. Int J Int Syst 20: 415–432, 2005.

Causal reasoning occupies a central position in human reasoning. Causality is granular in many ways. Knowledge of some causal effects is imprecise. Perhaps, complete knowledge of all possible factors might lead to crisp causal descriptions. However, it is unlikely that all possible factors can be known. Even if the precise elements are unknown, people recognize that a complex of elements can cause an effect. They may not know what events are in the complex; or, what constraints and laws impact the complex. Common sense understanding accepts imprecision, uncertainty and imperfect knowledge and is more successful reasoning with a few large-grain sized events than many fine-grained events. Perhaps, a satisfying solution would be to develop large-grained solutions and only go to an implicitly nested finer-grain when the impreciseness of the large-grain is unsatisfactory. Fuzzy Markov models might be used. It may be more computationally feasible to work on larger-grained representations.

Association rules are a data mining cornerstone. Association rules greatest impact is in helping to make decisions. One association rule quality measure is a rule's decision value. Association rules are often constructed using simplifying assumptions that lead to naive results and consequently naive decisions. Perhaps the greatest impact on decision value quality comes from treating association rules as causal statements without understanding whether there is, in fact, underlying causality. Complete knowledge of all possible factors might lead to a crisp description of whether an effect will occur. However, it is unlikely that all possible factors can be known. Commonsense world understanding accepts imprecision, uncertainty and imperfect knowledge. People recognize that a complex collection of elements can cause a particular effect. The events in the complex may not be known; or, what constraints and laws the complex is subject to. Sometimes, the details underlying an event can be known to a fine level of detail, sometimes not. Usually, commonsense reasoning is more successful in reasoning about a few large-grain sized events than many fine-grained events. A satisficing solution would be to develop large-grained solutions and only use the flner- grain when the impreciseness of the large-grain is unsatisfactory.

Determining causality has been a tantalizing goal throughout human history. Proper sacrifices to the gods were thought to bring rewards; failure to make suitable observations were thought to lead to disaster. Today, data mining holds the promise of extracting unsuspected information from very large databases. Methods have been developed to build association rules from large data sets. Association rules indicate the strength of association of two or more data attributes. In many ways, the interest in association rules is that they offer the promise (or illusion) of causal, or at least, predictive relationships. However, association rules only calculate a joint probability; they do not express a causal relationship. If causal relationships could be discovered, it would be very useful. Our goal is to explore causality in the data mining context.

Clustering groups items together that are most similar to each other and sets those that are least similar into different clusters. Methods have been developed to cluster records in a data set that are of only qualitative or quantitative data. Data sets exist that contain a mix of qualitative (nominal and ordinal) and quantitative (discrete and continuous) data. Clustering records of mixed kinds of data is a difficult problem. A metric to measure the similarity between records of mixed data types is needed. Once a clustering is found, we do not know how to best evaluate the quality of the clustering when there is a mixture of data varieties.

Causality occupies a position of centrality in human reasoning. It plays an essential role in commonsense human decision-making. Determining causes has been a tantalizing goal throughout human history. Proper sacrifices to the gods were thought to bring rewards; failure to make the proper observations to led to disaster. Today, data mining holds the promise of extracting unsuspected information from very large databases. The most common methods build association rules. In many ways, the interest in association rules is that they offer the promise (or illusion) of causal, or at least, predictive relationships. However, association rules only calculate a joint occurrence frequency; they do not express a causal relationship. If causal relationships could be discovered, it would be very useful. This paper explores the possible representation of causality drawn from large data sets.

Data mining discovers interesting information from a data set. Mining incorporates different methods and considers different kinds of information. Granulation is an important aspect of mining. The data sets can be extremely large with multiple kinds of data in high dimensionality. Without granulation, large data sets often are computationally infeasible; and, the generated results may be overly fine grained. Most available algorithms work with quantitative data. However, many data sets contain a mixture of quantitative and qualitative data. Our goal is to group records containing multiple data varieties: quantitative (discrete, continuous) and qualitative (ordinal, nominal). Grouping based on different quantitative metrics can be difficult. Incorporating various qualitative elements is not simple. There are partially successful strategies as well as several differential geometries. We expect to use a mixture of scalar methods and soft computing methods (rough sets, fuzzy sets), as well as methods using other metrics. To cluster whole records in a data set, it would be useful to have a general similarity metric or a set of integrated similarity metrics that would allow record to record similarity comparisons. There are methods to granulate data items belonging to a single attribute. Few methods exist that might meaningfully handle a combination of many data varieties in a single metric. This paper is an initial consideration of strategies for integrating multiple metrics in the task of granulating records.

Data mining holds the promise of extracting unsuspected information from very large databases: Methods have been developed to build association rules from categorical data. However, often many fine grained rules are generated. Additionally, much real data is not only categorical; it is quantitative. In forming association rules, quantitative values are often reduced to categorical values; this may overly simplify results. The concern of the work presented is considering how fine grained rules might be aggregated and the role that noncategorical data might have. It appears that soft computing techniques may be useful.

A computational approach is shown for unsupervised, reactive, database mining. This approach is dependent on soft computing techniques. Database mining seeks to discover noteworthy, unrecognized associations between database items. A novel approach is suggested for unsupervised search controlled by dissonance reduction. Both crisp and noncrisp data are subject to discovery. Another aspect of uncertainty is the metric that controls discovery. Issues involve: coherence measures, granularization, user intelligible results, unsupervised recognition of interesting results, and concept equivalent formation. © 1997 John Wiley & Sons, Inc.

An experiment was conducted that used a knowledge-based system to classify a set of natural language documents. An important issue in the design of knowledge-based systems is how to equip them with the capability to transmit uncertainty from imprecise premises to a conclusion and associate the conclusion with a level of belief. A study was carried out to compare the effectiveness of various fuzzy calculi. The test domain was full-text retrieval of natural language documents. This test domain was chosen because natural language is inherently imprecise and there are well-defined performance measures that can provide the basis of calculi comparison. Both robustness and relative retrieval quality were studied. Some calculi were found to be more effective. The most commonly used knowledge-based systems calculi were not the best performing.>

We designed and executed a course to develop expert system expertise in a classroom environment. It taught both the theory and practice of knowledge-based systems. Teams consisted of knowledge domain experts and computer experienced people. The interest was training people in knowledge-base tasks and having them develop commercially significant projects. The focust of this paper is on what was done in the classroom to provide a significant expert system development experience.

This is a theoretical expositional exploration into the underlying needs of concept formation. The main purpose is to identify and discuss the differing forms of categorization in the context of possible machine learning and representation of those concepts. Conceptualization is the process of developing the abstractions that are needed to support reasoning. The formation of the machine equivalent of human concepts is critical to the development of a general, machine based reasoning capacity. An aid in understanding conceptual categorization is prototype theory. It helps to identify the building block tools that are used to construct categories and taxonomies. When developing taxonomic structures, framing conflicts can occur in terms of how things should be clustered together, what should be the relative hierarchal levels, and what should be subordinate to what.

The usefulness of introducing subprograms (PROCEDUREs and FUNCTIONs) as the first program control structure in an introductory programming course is discussed. The motivation for an instructor to do this is to place an earlier and greater emphasis on top-down design and structured programming. Specific pedalogical examples are provided.

The best structure for a first course in software engineering is unclear. First, because what should be taught has not been firmly established to the general satisfaction of those involved. Second, providing a realistic environment to motivate belief in the utility of the practices involved is difficult.

Unidirectional trees and lattices may both be used to hold a symbolic data base consisting of lexes, lexemes or other symbol strings. This paper empirically compares placing symbolic information into both trees and lattices (a lattice may be thought of as a unidirectional network with single and initiating and terminating nodes).

L. J. Mazlack; Predicting student success in an introductory programming course, The Computer Journal, Volume 21, Issue 4, 1 January 1978, Pages 380–384, h

There are three different approaches to an introductory computer science course: technical competance, non-technical awareness, and a mixture of technical competance and non-technical awareness. This paper discusses the best strategy for a course aimed at students who will not need a high level of technical competance once they leave the course. A course which develops a computer awareness through a mixture of technical programming instruction and a discussion of computer applications and power for good or evil. Films are extensively used.

This paper reports on the construction of a crossword puzzle generator. After an unsuccessful attempt to construct puzzles by whole word insertion, puzzles were constructed letter by letter. Heuristically determined decision structure was required. The constructor resolved questions of letter selection, ordering and reordering of the solution sequence, dictionary structure and access, and decision path selection. The decision basis for letter selection was based on a pseudo-probabilistic approach.