Back to search

National and Kapodistrian University of Athens

Μελέτη μεθοδολογιών παραλληλισμού για την υλισμική απεικόνιση της άλγεβρας δυαδικών σχέσεων και κανόνων από δένδρα αποφάσεων

Abstract

dc:description

During the operation of an object oriented information system, the objects derived are holding references one of another. These objects can be viewed as interconnected forming relationships amongst them. A set of interconnected objects can also be considered as a set of pairs of objects, since the pairing implies a relation (connection) between the pair members. A binary relation is defined as a set of such object pairs. The algebra of binary relations (Tarski 1941, Chin 1951, Jonsson 1951, Jonsson 1952, Tarksi 1987, Givant 1994) can provide an algebraic framework (Sarathy 1993, Gyssens 1994a, Van den Bussche 2001) for manipulating object oriented databases. Previous attempts to use the algebra of binary relations for object manipulation purposes resulted in graph-oriented systems, where graphs were used for querying, browsing, updating and restructuring object databases (Gyssens 1990, Gyssens 1994b, Sarathy 1993) or architectures of software systems (Hold 1998). In this dissertation, instead of using graphs, an extension to the OQL of the ODMG’s object data standard 3.0 for incorporating binary relational expressions for query declarations is proposed. In this proposal, the introduction of simple and multiple merged chains of binary relations in query declarations extends the concept of relation composition, which is defined by the binary relational algebra. A Semi- Dewavefront-like Algorithm, which is mapped on a Bidirectional Multi- Functional Engine-BMFE is also introduced for efficient query execution purposes. This algorithm (and engine) processes the simple and multiple merged chains of binary relations introduced in the OQL extension and declared in queries. The basic elements of the architecture of this engine can be of two types of array processors, a synchronous Square Systolic Device (for Wavefront-like Computations) - S2DWC and an asynchronous Bi-directional Parallel Cubic Engine - BPCE. These array processors are the mappings of the semi-group structure of the binary relational algebra. The multifunctional engine is an extension of this mapping, since it processes, across multiple directions, multiple relations structured as simple or multiple merged chains. These systems were designed and extensively tested using environments provided by suitable VHDL tools and were FPGA implemented virtually producing hybrid systems, composed by both software and hardware (middleware). Decision trees or rules that are derived from machine learning algorithms can be further processed, especially if they are of special forms that serve concrete needs. Data accessing via decision trees is also investigated in this dissertation, through a proposed algorithm for the extraction of rules from decision trees with complex nominal data that could be even in the form of free text. The goal of the algorithm is to process a decision tree generated using the classification algorithm ID3, which processed data in natural language expressions form and to extract rules expressed in natural language. In this case, the algorithm’s input tree is full of complex nominal expressions. Since this tree is of unknown form and size, it cannot be easily processed using parallel techniques. However, it does not apply for the hardware mapping of the rules themselves. Herein are investigated certain parallel methodologies which are applied in the design and implementation of a chip that is able to recognize and classify patterns of medical measures relative to the breast cancer disease. This chip called Integrated Intelligent Medical Classification Engine - I2MCE was FPGA implemented and extensively tested in a VHDL environment. Measures taken from the patient are fed into this chip that recognizes them according to the classification rules incorporated in its design. The result achieved concerns the nature of the cancer under examination, namely, benign or malignant. A parallel program that simulates the medical diagnostic chip exhibits the parallelism of these methodologies. The program is implemented using the threads and the inter-thread communication of the java programming language. A series of similar medical diagnostic chips can be designed and implemented for a number of diseases using the introduced methodologies. These chips can be incorporated in portable microdevices that can be used by doctors and health care professionals in general for diagnosis decision support purposes.

Degree

thesis:*
Grantor dc:publisher
National and Kapodistrian University of Athens
Year dc:date
2005

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Fountoukis, Spyridon
  • Φουντούκης, Σπυρίδων

Subjects

dc:subject × 20

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/21344

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Fountoukis, Spyridon; Φουντούκης, Σπυρίδων. Μελέτη μεθοδολογιών παραλληλισμού για την υλισμική απεικόνιση της άλγεβρας δυαδικών σχέσεων και κανόνων από δένδρα αποφάσεων. National and Kapodistrian University of Athens, 2005. http://hdl.handle.net/10442/hedi/21344