Agent technology Enabling communication among tools and data

1
Agent technology Enabling communication among
tools and data

• Thomas E. Potok
• Collaborative Technologies Group Leader
• Computer Science and Mathematics Division
• Oak Ridge National Laboratory

Mladen Vouk Professor of Computer Science
Technical Director of the NC State Center for
Advanced Computing and Communications North
Carolina State University
2
Agent Technology Trend
Tim Berners-Lee The Semantic Web will open up
the knowledge and workings of humankind to
meaningful analysis by software agents, providing
a new class of tools by which we can live, work
and learn together The Semantic Web Scientific
American 5/2001
3
Successful Projects
VIPAR Knowledge Discovery

• We have extensive expertise in agent development
• Began working with agent technologies in 1980s

Supply Chain Management Agent System
Manufacturing Emulation Agent System
Collaborative Decision Support System
Neural Nets for Recovery Boiler Control
Neural Nets for Bankruptcy Prediction
Neural Nets for Spring-back Prediction
Collaborative Design System
Neural Nets for Resistance. Spot Welding
Neural Nets for Material Mix Optimization
Genetic Algorithms for Chemical Synthesis

• Over 10 successful projects within the last 5
  years
• Collaborations with leading agent experts

Knowledge-based Systems - Manufacturing Advisors
Knowledge-based Systems for Constructability
Design and Analysis of Computer Experiments
Knowledge-based Computer Systems Calibration
1985
1990
1995
2000
4
What are Agents?
...Software entities that assist people and act
on their behalf - IBM ...Software robots
Proactive detect changes in their environment and
react to those in a timely manner by answering
to events and initiating actions
Traditional Software Object
Goal-driven have a purpose and act in accordance
with that purpose until it is fulfilled
Behavior
State
Agent Technology
Communicative able to interact and communicate
with users and other agents
Autonomous can have control over their own
actions and be able to work and launch actions
independent of the user or other actors
Learning have the ability to learn from
experiences in their environment
5
Typical Agent Architecture
MULTIAGENT APPLICATION (Agent-User Interface,
Conversation Interface)
SOCIAL MODEL (Interaction rules, Conversation
Management)
Legacy Databases
LOCAL MODEL (Domain KB, Ontologies)
INFERENCE ENGINE (Decision Support Models)
Legacy Software
LINGUISTIC LAYER (Message Syntax and Semantics)
COMMUNICATION PROTOCOL LAYER (Multicast, Directed
Communication)
LEARNING LAYER
INFRASTRUCTURE (e.g., Java Virtual Machine)
OPERATING SYSTEM
HARDWARE
6
Current Agent Funding

• NSF
• Division of Information and Intelligent Systems
• Division of Electrical and Communications Systems
  
• Computation and Social Systems
• Knowledge and Cognitive Systems
• Control, Networks, and Computational Intelligence
  (CNCI)
• DARPA
• Information Technology Office
• Control of Agent-Based Systems (CoABS)
• DARPA Agent Mark Up Language (DAML)
• Taskable Agent Software Kit (TASK)
• Office of Naval Research
• Information, Electronics Surveillance
• Intelligent Systems
• Others include
• NASA, Army, Air force, Office of the Secretary of
  Defense, NIST,

7
Simple Agent Example
Agent, find me the book War and Peace, and I
need it tomorrow
Amazon
2 Days 18.50
Dedicated Agents
Barnes and Nobel
1 Day 21.75
B. Dalton
1 Day 20.25
Agent Communities
Library
1 Day Free
Does the agent understand buying books?
Form a plan to buy the book
eBay
1 Day 12.50
Execute the plan
8
VIPAR Project US Pacific Command

• Analysts Search 20-30 papers a day
• Manually find the articles on internet
• Manually organizing the articles

Unorganized Newspaper Articles
Analysts
Organized Results
Inside China
NEA
Russia Today
SEA
Jakarta Post
Oceania
9
New Intelligent Agent Approach
Intelligent Agent
Inside China
Analysts
Economy
Russia Today
Military
Jakarta Post
Political
Retrieval Agents
Vector Perturbation Clustering
Knowledge Bins
10
Oak Ridge Mobile Agent Community (ORMAC)
Communication
Machine A
Machine B
Agent Host X
Agent Host Y
Agent Host Z
Agent Context
Agent Context
Agent Context
Agent 1
Agent 2
Agent 3

• Allows agents to be transmitted and received
  among machines.
• Agents can also interact with systems and agents
  that are not part of the community.
• Agents uses the Foundation for Intelligent
  Physical Agent (FIPA) compliant agent control
  language (ACL) messages.

11
VIPAR Architecture
RDF Ontologies
Dynamic Vector Space Models
Conversion from HTML to XML
Multi-agent System Built on ORMAC Framework
12
Clustering A Simple Example
2. Terms Dictionary
1. Original and Parsed Documents
3. Vector Space Model
4. Document Similarity
5. Euclidean Distance
7. Phylips tree representation
6. Agglomerative Hierarchical Clustering
Agglomerative
Divisive
8
1
2
3
4
6
7
13
VIPAR Knowledge Management Same result, easier to
see

• Wen Ho Lee
• Wen Ho Lee Spends First Day Savoring Home
  Delights
• Clinton Calls For Review Of Lee Secrets Case
• Clinton Concerned Over Lee Case - Reno On
  Defensive
• Asian-Americans Demand White House Action on Lee
• Lee Case Points up Scientists' Attitude on
  Security
• India and Pakistan
• India and Pakistan Troubled relations
• Troops die in Kashmir clashes
• IAEA Meeting
• IAEA Supports Putin Nuclear Power Initiative
• China Rejects Moves to Tighten Regulation of
  Nuclear Materials
• U.S. China Trade
• U.S. China Trade Vote Milestone on Rocky Road

2
3
1
4
Wen Ho Lee
5
US China Trade
India/Pakistan
6
10
IAEA Meeting
7
9
8
14
Endorsements

• US Pacific Commands CINC Admiral Blair
• A tremendously successful project
• Software agents lead to substantially improved
  analytical products.
• US Pacific Commands Science and Technology
  Advisor
• hit a grand slam home run!
• first time we've seen information discovery and
  knowledge management software working here at HQ
  USCINCPAC operationally.

15
How agents can work for scientific data
16
Astrophysics simulation of temperature and
tangential velocities
Visualization created by Ross Toedte
17
Hypothetical Scenario
Probe
Data Reduction Agent
Clustering Agent
Parallel IO Agent
Astrophysics Simulation
Dynamic Clustering
Stripe Analysis
Binning Sampling
Astrophysics Data Agent
MPI-IO
HPSS
Storage Agent
Full Data
Reduced Data
Clustered Data
18
Example
SDM Agent Community
Scientist What if we reduce the data?
A5
A1
A2
A3
A4
Agent whiteboard
A1 REQUEST All Dimensionality Reduction FORMAT
HDF5 XML TYPE Climate A2 REPLY A1
Dimensionality Reduction FORMAT HDF5 TYPE
General A3 REPLY A1 Dimensionality Reduction
FORMAT ASCII TYPE Physics A1 REQUEST All
Conversion FORMAT HDF5 XML to HDF5 TYPE any A4
REPLY A1 Conversion FORMAT HDF5 XML to HDF5
TYPE Nuclear A5 REPLY A1 Conversion FORMAT
HDF5 XML to HDF4 XML TYPE any A1 PERFORM A4
Conversion FORMAT HDF5 XML to HDF5 DATA D1 A4
RESULT A1 Conversion FORMAT HDF5 XML to HDF5
Data R1 A1 PERFORM A2 Dimensionality Reduction
FORMAT HDF5 Data R1 A2 RESULTS A1
Dimensionality Reduction FORMAT HDF5 Data
R2
19
Year 1 Deliverables

• Generic scientific data ontology
• Current framework supports a standard FIPA
  message ACL
• Need for agents to be able understand and
  interact with scientific data
• Data centric?
• Tool centric?
• Analysis centric?

20
Year 1 Deliverables continued

• Generic data and model agents for data mining and
  discovery of access patterns
• Define general case agents that can be adapted to
  work with
• Specific data types
• Specific tools
• Establish network-based software configuration,
  management and distribution site

21
Summary

• Intelligent agents are a significant future
  technology direction
• We have a significant progression of deployed
  agent systems and research to our credit
• We have ringing endorsements regarding the
  success of our systems and research
• We look forward to the next steps

22
(No Transcript)
23
How to use agents?

• We do not what to redo what you have already
  done!!!
• We want to supply the data communication
  protocols so that scientists can use your tools
  more effectively
• Java based, FIPA compliant