Parallel and Distributed Intelligent Systems
Download
Report
Transcript Parallel and Distributed Intelligent Systems
Parallel and Distributed Intelligent
Systems
Virendrakumar C. Bhavsar
Professor and
Director, Advanced Computational Research Laboratory
Faculty of Computer Science
University of New Brunswick Fredericton, NB
[email protected]
www.cs.unb.ca/profs/bhavsar
www.cs.unb.ca/acrl
Outline
Past Research Work
Current Research Work
Conclusion
Past Research Work
Parallel/Distributed Processing
- Parallel Computer Architecture
- Design and Analysis of Parallel Algorithms
- Real-time and Fault-Tolerant Systems
Artificial Neural Networks
Learning Machines and Evolutionary
Computation
Computer Graphics
Visualization
Advanced Computational Research
Laboratory
High Performance Computational ProblemSolving and Visualization Environment
Computational Experiments in multiple
disciplines: CS, Science and Eng.
16-Processor IBM SP3
Member of C3.ca Association, Inc.
(http://www.c3.ca)
Advanced Computational Research
Laboratory
www.cs.unb.ca/acrl
Virendra Bhavsar, Director
Chris MacPhee, Scientific Computing Support
Sean Seeley, System Administrator
ACRL’s IBM SP
– 16 processors; 24
Gigabit
Ethernet
GFLOPS
Disk
4 Winterhawk II nodes
High Perforrnance
Switch
IBM SP at ACRL: The Clustered SMP
Four 4-way SMPs
Each node has its own copy
of the O/S
Processors on the node are
closer than those on different
nodes
IBM Power3 SP Switch
Bidirectional multistage
interconnection networks
(MIN)
300 MB/sec bi-directional
1.2 sec latency
Past Research Work (cont.)
Multimedia for Education: Intelligent Tutoring
Systems
Multi-Lingual Systems and Transliteration
Web Portal with an Intelligent User Profile
Generator
Multi-Agent Systems
Supervision/Co-supervision
50 master's theses; 4 doctoral theses
5 post-doctoral fellows/research associates
Current Research Work
Parallel/Distributed Processing
-PaGrid: A Mesh Partitioner for Computational Grids
- Dynamic Partitioning for Efficient Processing on Parallel
Computers
Multi-Agent Systems (Distributed Artificial Intelligence)
- Multi-Agent System for Automatic Annotation of EST
Sequences (funded by ‘The Canadian Potato Genomics’)
- CS6999: Multi-Agent Systems
- Dynamic Clustering of Agents in the Café
- Agents with Ontology-based Keyphrases and Tree-distance
algorithms
- Scalability studies of Multi-Agent Systems
- eCommerce applications
Current Research Work
eLearning (eduSorceCanada Project)
- Reuse and exchange course content stored as “learning
objects.’’
- Implementation and testing of learning objects using CanCore
metadata
-XML schema for content packaging
- other projects
What is a GRID System
Cooperative network of shared resources
- Includes computers, network links, human resources and
databases
Supports the development of advanced R&D
applications in Science, Engineering and
Technology Development, Finance and the Arts.
March, 2000
Copyright (C) C3.ca
15
GRID Applications
Large scale and resource intensive frontier applications
– R&D applications that go beyond current technological
capabilities
– Technology development applications in multi-media,
finance, production arts, hard sciences and engineering.
- Multi-media applications such as embedded video,
digital video servers and video conferencing.
March, 2000
Copyright (C) C3.ca
16
Current C3.ca RP Network
March, 2000
Copyright (C) C3.ca
17
The Canadian Potato Genomics Project
ATLANTIC
CANADA
• 46% of national
potato production
$1 Billion/year
• Home of McCain
Foods Ltd.
$5.5 billion/year
• Potato Research
Center of AAFC
• Solanum Genomics
International Inc.
The Canadian Potato Genomics Project
Research Areas
• Bioinformatic Analysis
• Access to resources via CBR membership/node status
• Raw sequence processing and analysis by Fredericton
bioinformatics group
(Vector trimming, base calling, clustering, contig assembly, BLAST, an
• Relational database management system of CPGP to
link NRC (sequencing), CBR and researchers
• In silico assignment of gene function
• Microarray data
The Canadian Potato Genomics Project
Research Areas
• Bioinformatic Research To Suit Project Needs (UNB):
• Autonomous agent development to automatically update
sequence annotations
• Enhancement of bioinformatic algorithm performance
with parallel computing
• Algorithm development using annotation information to
enhance sequence searching
• The application of clustering and learning techniques to
the analysis of expression data
[email protected]
[email protected]
ucsd.edu
ai.it.nrc.ca
S
e
r
v
e
r
S
e
r
v
e
r
[email protected]
[email protected]
[email protected]
Café
anwhere.else
[email protected]
Café
S
e
r
v
e
r
cs.stir.ac.uk
meto.gov.uk
S
e
r
v
e
r
S
e
r
v
e
r
Café
Clients
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
Performance Evaluation of ACORN
Test-bed: Several Autonomous Servers, each
serving autonomous virtual users
Virtual User - capable of creating agents
- picks up a topic from a client core’s interest
- migrates to other servers
- potential destinations
Performance Evaluation of ACORN
120
Time (Seconds)
100
80
60
40
20
0
0
0
20
40
60
Number of Agents
80
Mingle Tim e with cosine
Total Exection Time with cosine
Mingle Tim e for without cosine
Total Exection Time without cosine
100
Why learning objects?
• COST: 1000s of colleges have common course topics
• large numbers of courses are going online
• World does not need 1000s of similar learning topics
• World needs only about a dozen
• Expensive to develop so sharing is essential
(From Downes, 2000)
What is METADATA?
data about data
Metadata standards are agreed-on criteria for
describing data to support interoperability
Example:
January 31, 2001
31 janvier 2001
2001-01-31
01-31-2000
31012000
Metadata and RDF
implementation
* XML
* Resource Description Framework
(RDF) = structure
Metadata =
semantics & resources
Conclusion
Parallel/Distributed Processing
Multi-Agent Systems (Distributed Artificial Intelligence)
NSERC Project, The Canadian Potato Genomics Project
eLearning (eduSorceCanada Project)
Automated and manually-driven user profile generation
and update