Physical Sciences and Mathematics Commons^™

A Visualization System For Information Retrieval And Mining In High Dimensional Databases, Xinhuan Zheng

Theses

In this thesis, we present a search engine capable of giving good heuristic answers to the queries on a structural database. A structural database holds structural objects, e.g., protein secondary and tertiary structures, 3D molecules, phylogenetic trees, neuroanatomical networks, parse trees, CAD/CAM parts, and XML documents. Answering queries on such databases often requires solving variants of the graph isomorphism or subisomorphism problems. We also describe a graphic user interface which interacts with users to facilitate visualizing query results. We use 3D molecules (graphs) as illustrating examples, though our prototype is able to handle many other different types of structural data.

Analysis Of Clustering Algorithms For Spike Sorting Of Multiunit Extracellular Recordings, Jayesh Rege

Theses

Various techniques have been considered in the past to identify distinct spike shapes from mulitunit extracellular recording. These techniques involve adaptive filtering techniques or template matching techniques or hierarchical clustering techniques. In this investigation, we have used Principal Component Analysis followed by various clustering techniques to identify distinct spike shapes. The amplitude filter is used to separate spikes from background neuronal activity. The correlation matrix of the spike data is used to compute principal component wave forms. Each spike is thus represented by the coefficients of principal components. Then, We have used agglomorative hierarchical clustering algorithm to perform the initial …

Implementation Of A Geometric Hashing Technique And It's Application To 3d Molecular Structure Search, Jennifer Lynn Cerequas

Theses

This paper proposes the application of a geometric hash technique to the searching of 3D chemical structures. Chemical structures are represented in global XYZ coordinate format. An algorithm is applied to the substructures within the existing chemical structures to hash them into hash tables on disk. A query substructure is then hashed to find matches (hits) of the existing hash tables. The entries in the matching hash tables are compared to the query substructure to find the existing substructures that are an "approximate" match.

The result is a technique which allows existing substructures to be compared to a (new) query …

Flag : The Fault-Line Analytic Graph And Fingerprint Classification, Ching-Yu Huang

Dissertations

Fingerprints can be classified into millions of groups by quantitative measurements of their new representations - Fault-Line Analytic Graphs (FLAG), which describe the relationship between ridge flows and singular points. This new model is highly mathematical, therefore, human interpretation can be reduced to a minimum and the time of identification can be significantly reduced.

There are some well known features on fingerprints such as singular points, cores and deltas, which are global features which characterize the fingerprint pattern class, and minutiae which are the local features which characterize an individual fingerprint image. Singular points are more important than minutiae when …

A Simple Neural Agent Communicating Through Sets, James P. Stanski

Theses

Networked agents of the simplest kind offer the power of cooperative problem solving through parallel operation along with tight packaging potential. Such agents are self-contained analog machines capable of only a few primitive intelligent operations. In this thesis, a design will be developed for a simple agent capable of sending, receiving, and processing information in a environment where agents are coupled together. This environment imposes unorchestrated simultaneous input while expecting a useful timely response. Successful collaboration in these conditions is accomplished through sets encoded within pulse ensembles. The simplicity of the set definition is an inviting candidate for message communication …

Pattern Discovery In Sequence Databases : Algorithms And Applications To Dna/Protein Classification, Gung-Wei Chirn

Dissertations

Sequence databases comprise sequence data, which are linear structural descriptions of many natural entities. Approximate pattern discovery in a sequence database can lead to important conclusions or prediction of new phenomena. Traditional database technology is not suitable for accomplishing the task, and new techniques need to be developed.

In this dissertation, we propose several new techniques for discovering patterns in sequence databases. Our techniques incorporate pattern matching algorithms and novel heuristics for discovery and optimization. Experimental results of applying the techniques to both generated data and DNA/proteins show the effectiveness of the proposed techniques.

We then develop several classifiers using …

Application Of Approximate Graph Matching Techniques For Searching Databases Of Two-Dimensional Chemical Structures, Karen R. Pysniak

Theses

This paper proposes the application of approximate graph matching techniques for best-match searching of two-dimensional chemical structure databases based upon topology. Chemical structures are represented as labeled graphs, each atom a node in the graph and each bond an edge. By inserting; deleting and renaming nodes/edges, one structure may be transformed into another. We define similarity as the weighted sum of the costs of these edit operations. An algorithm for approximating the minimum distance between two graphs based on simulated annealing is applied. Best-match searches are performed utilizing this pre-computed distance information and applying the concepts of triangle inequality to …