Raj Bhatnagar

Handling Uncertain Information: A Review of Numeric and Non-numeric Methods

Problem solving and decision making by humans is often done in environments where information concerning the problem is partial or approximate. AI researchers have been attempting to emulate this capability in computer expert systems. Most of the methods used to-date lack a theoretical foundation. Some theories for handling uncertainty of information have been proposed in the recent past. In this paper, we critically review these theories. The main theories that we examine are: Probability Theory, Shafers Evidence Theory, Zadehs Possibility Theory, Cohens Theory of Endorsements and the non-monotonic logics. We describe these in terms of the representation of uncertain information, and combination of bodies of information and inferencing with such information, and consider the strong and weak aspects of each theory.

Similarity measures in formal concept analysis

Formal concept analysis (FCA) has been applied successively in diverse fields such as data mining, conceptual modeling, social networks, software engineering, and the semantic web. One shortcoming of FCA, however, is the large number of concepts that typically arise in dense datasets hindering typical tasks such as rule generation and visualization. To overcome this shortcoming, it is important to develop formalisms and methods to segment, categorize and cluster formal concepts. The first step in achieving these aims is to define suitable similarity and dissimilarity measures of formal concepts. In this paper we propose three similarity measures based on existent set-based measures in addition to developing the completely novel zeros-induced measure. Moreover, we formally prove that all the measures proposed are indeed similarity measures and investigate the computational complexity of computing them. Finally, an extensive empirical evaluation on real-world data is presented in which the utility and character of each similarity measure is tested and evaluated.

Prediction and Analysis of the Protein Interactome in Pseudomonas aeruginosa to Enable Network-Based Drug Target Selection

Pseudomonas aeruginosa (PA) is a ubiquitous opportunistic pathogen that is capable of causing highly problematic, chronic infections in cystic fibrosis and chronic obstructive pulmonary disease patients. With the increased prevalence of multi-drug resistant PA, the conventional “one gene, one drug, one disease” paradigm is losing effectiveness. Network pharmacology, on the other hand, may hold the promise of discovering new drug targets to treat a variety of PA infections. However, given the urgent need for novel drug target discovery, a PA protein-protein interaction (PPI) network of high accuracy and coverage, has not yet been constructed. In this study, we predicted a genome-scale PPI network of PA by integrating various genomic features of PA proteins/genes by a machine learning-based approach. A total of 54,107 interactions covering 4,181 proteins in PA were predicted. A high-confidence network combining predicted high-confidence interactions, a reference set and verified interactions that consist of 3,343 proteins and 19,416 potential interactions was further assembled and analyzed. The predicted interactome network from this study is the first large-scale PPI network in PA with significant coverage and high accuracy. Subsequent analysis, including validations based on existing small-scale PPI data and the network structure comparison with other model organisms, shows the validity of the predicted PPI network. Potential drug targets were identified and prioritized based on their essentiality and topological importance in the high-confidence network. Host-pathogen protein interactions between human and PA were further extracted and analyzed. In addition, case studies were performed on protein interactions regarding anti-sigma factor MucA, negative periplasmic alginate regulator MucB, and the transcriptional regulator RhlR. A web server to access the predicted PPI dataset is available at http://research.cchmc.org/PPIdatabase/.

Development associated profiling of chitinase and microRNA of Helicoverpa armigera identified chitinase repressive microRNA

Expression of chitinase is developmentally regulated in insects in consonance with their molting process. During the larval-larval metamorphosis in Helicoverpa armigera, chitinase gene expression varies from high to negligible. In the five-day metamorphic course of fifth-instar larvae, chitinase transcript is least abundant on third day and maximal on fifth day. MicroRNA library prepared from these highest and lowest chitinase-expressing larval stages resulted in isolation of several miRNAs. In silico analysis of sequenced miRNAs revealed three miRNAs having sequence similarity to 3′UTR of chitinase. Gene-targeted specific action of these miRNAs, was investigated by luciferase reporter having 3′UTR of chitinase. Only one of three miRNAs, miR-24, inhibited luciferase expression. Further, a day-wise in vivo quantification of miR-24 in fifth-instar larvae revealed a negative correlation with corresponding chitinase transcript abundance. The force-feeding of synthetic miR-24 induced significant morphological aberrations accompanied with arrest of molting. These miR-24 force-fed larvae revealed significantly reduced chitinase transcript abundance.

Algorithm for Discovering Low-Variance 3-Clusters from Real-Valued Datasets

The concept of Triclusters has been investigated recently in the context of two relational datasets that share labels along one of the dimensions. By simultaneously processing two datasets to unveil triclusters, new useful knowledge and insights can be obtained. However, some recently reported methods are either closely linked to specific problems or constrain datasets to have some specific distributions. Algorithms for generating triclusters whose cell-values demonstrate simple well known statistical properties, such as upper bounds on standard deviations, are needed for many applications. In this paper we present a 3-Clustering algorithm that searches for meaningful combinations of biclusters in two related datasets. The algorithm can handle situations involving: (i) datasets in which a few data objects may be present in only one dataset and not in both datasets, (ii) the two datasets may have different numbers of objects and/or attributes, and (iii) the cell-value distributions in two datasets may be different. In our formulation the cell-values of each selected tricluster, formed by two independent biclusters, are such that the standard deviations in each bicluster obeys an upper bound and the sets of objects in the two biclusters overlap to the maximum possible extent. We present validation of our algorithm by presenting the properties of the 3-Clusters discovered from a synthetic dataset and from a real world cross-species genomic dataset. The results of our algorithm unveil interesting insights for the cross-species genomic domain.

An Efficient Constraint-Based Closed Set Mining Algorithm

We present a search algorithm for mining closed sets in high dimensional binary datasets. Our algorithm is designed for dense datasets, where the percentage of 1s in the dataset is usually higher than 10%, and the total number of closed sets is much larger than the number of objects in the dataset. Our algorithm is memory efficient since, unlike many other closed set mining algorithms, it does not require all patterns mined so far to be kept in the memory. Optimization techniques are introduced in this paper, and we also present a parallel version of our algorithm.

Discovery of Temporal Dependencies between Frequent Patterns in Multivariate Time Series

We consider the problem of mining multivariate time series data for discovering (i) frequently occurring substring patterns in a dimension, (ii) temporal associations among these substring patterns within or across different dimensions, and (iii) large intervals that sustain a particular mode of operation. These represent patterns at three different levels of abstraction for a dataset having very fine granularity. Discovery of such temporal associations in a multivariate setting provides useful insights which results in a prediction and diagnostic capability for the domain. In this paper we present a methodology for efficiently discovering all frequent patterns in each dimension of the data using Suffix Trees; then clustering these substring patterns to construct equivalence classes of similar (approximately matching) patterns; and then searching for temporal dependencies among these equivalence classes using an efficient search algorithm. Modes of operation are then inferred as summarization of these temporal dependencies. Our method is generalizable, scalable, and can be adapted to provide robustness against noise, shifting, and scaling factors

Structural and probabilistic knowledge for abductive reasoning

Different ways of representing probabilistic relationships among the attributes of a domain ar examined, and it is shown that the nature of domain relationships used in a representation affects the types of reasoning objectives that can be achieved. Two well-known formalisms for representing the probabilistic among attributes of a domain. These are the dependence tree formalism presented by C.K. Chow and C.N. Liu (1968) and the Bayesian networks methodology presented by J. Pearl (1986). An example is used to illustrate the nature of the relationships and the difference in the types of reasoning performed by these two representations. An abductive type of reasoning objective that requires use of the known qualitative relationships of the domain is demonstrated. A suitable way to represent such qualitative relationships along with the probabilistic knowledge is given, and how an explanation for a set of observed events may be constituted is discussed. An algorithm for learning the qualitative relationships from empirical data using an algorithm based on the minimization of conditional entropy is presented. >

An effective algorithm for mining 3-clusters in vertically partitioned data

Conventional clustering algorithms group similar data points together along one dimension of a data table. Bi-clustering simultaneously clusters both dimensions of a data table. 3-clustering goes one step further and aims to concurrently cluster two data tables that share a common set of row labels, but whose column labels are distinct. Such clusters reveal the underlying connections between the elements of all three sets. We present a novel algorithm that discovers 3-clusters across vertically partitioned data. Our approach presents two new and important formulations: first we introduce the notion of a 3-cluster in partitioned data; and second we present a mathematical formulation that measures the quality of such clusters. Our algorithm discovers high quality, arbitrarily positioned, overlapping clusters, and is efficient in time. These results are exhibited in a comprehensive study on real datasets.

Evaluation of properties in the transition of capability based agent organization

It has been said that the only constant in life is change. This rule can also be directly applied to the lives of organizations. Any organization of non-trivial size, scope, life expectancy or function, is destined to change. An organization without the ability to transition is not robust, evolvable or adaptable within its environment. These basic preconditions to human organizations must also hold in viable agent organizations. To model an adaptable agent organization, the capability must be present to transition from one state to the next over the life of the organization. The organization model must include not only the structural objects, but also the ability to facilitate change. The ability to change empowers the organization to transition from one state to the next, over its useful life. To enable transition, we must formally capture and define what triggers an organization transition. In this paper, we will define the properties to formally model the ability of an adaptable organization to transition throughout its useful life. The properties will be instantiated, using an implemented system, allowing the evaluation of internal and external stimuli to cause transition to the organization. These transitions will be evaluated from several perspectives to determine their effectivity on basis of design and use.