Links

• http://brexperts.ning.com/events/brxgaa-february-skype-call-on (BRxG event)
• http://bizrules.info/network/brxg_20100216.htm (permalink)

1. ATTENDEES

• JM , Jason Morris , ORF2010 location options

• DH , Dave Holz

• SN , Steve Nunez

• AW , Andrew Waterman

• JF , Jacob Feldman

• MP , Mark Proctor

• PCL , Po-Chuan Lin

• RH , Rolando Hernandez

2. MEETING AGENDA

BRxG AA | Feb. 16, 2010  |  3rd Tuesday |  7:45 PM to 9:00PM CDT (Dallas)

7:45 - 8:00 PM - Call in, connect, and join the meeting

8:00 - 8:10 PM - Introduce new members; Discuss any new issues, challenges, problems & solutions, tips & ideas

8:10 - 9:00 PM - Present today's topics & questions for the experts; Expert discussion

9:00 PM - Wrap-up and suggestions for future topics

3. INTRODUCTIONS

• JM , ORF2010 location options ,
• SN - consider a rules conference in Europe…. Working on his benchmarks, converting to Jess, under new framework, blogs.illation.com.au
  • Lets Join him down under for a rules conf !? Volunteers?
  • Financial services
  • Sees lots of govt adopting BR… CUSTOMS, ANTI-TERRORISM INFO

• AW ,  JF , MF ,
• PCL - Java developer, beginner in rules; Has been to EVERY DRG meeting, starting class with James
• JCO

4. DRG/BRxG UPDATE
   1. DRG / JCO's group will meet in Dallas on 2nd Tuesday
      1. 7-8 PM Practical applications (i.e. practical, real-life or real hard business problem)
      2. 8-9 PM Theory (i.e. solution, design & Implementation)
      3. Future plans: Vendor presentations from IBM ILOG, FICO, PEGA, etc

5. OPEN DISCUSSION
   1. ASK THE EXPERT PANEL FOR THEIR DEFINITIONS of inference, chaining, forward/backward… and why use it… when do you need it?
      1. JM - traditional forward chaining supports deductive reasoning
         1. Backward chaining uses abductive(?) reasoning, to work form a premise or theory or hypothesis to look for a fact that supports that theory or hypotheesis
         2. There are cases where u use both, in the same app
         3. Desirable to use both types in an app
         4. Forward
            1. To rapidly get to a conclusion if you have facts readily available to chain thru
         5. Backward
            1. To pull facts into a program when u are missing data
            2. Many times see backward chaining to connect to db to get data and supply facts to an app
         6. When u combine these 2 types, u have powerful reasoning mechanisms
         7. He likes to do classic Expert Systems, vs the rest of us :)
            1. He uses rules as control mechanisms meta-rules to control flow & OVERSIGHT,
            2. Layers of rules that actively PROCESS , other rules that WATCH OVER health and status of the apps
         8. DH uses rules to dynamically gen the GUIS
            1. They declaratively control user interaction using rules
            2. They supply facts to control UI states
         9. Lots of different ways to use rules, not just the "popular" business rules approach
      2. MP
         1. Looks at it from a diff point of view, he teaches java developers
            1. Forward - analogy DB and SQL - i.e. a rule is like a trigger
               1. ____ is like a consequence (THEN)
               2. Materialized view with a trigger
            2. Backward - for developers to understand…
               1. Look at opportunistic backward chaining
                  1. He means lazy field / lazy object optimization
               2. Similar to QUERIES , i.e. corrolated sub-queries
            3. Also, populating data from a source
               1. "materializing" ??? - std oracle feature, materialized view
                  1. SELECT * FROM VIEW
                  2. Materialized view that can have triggers on it
            4. "A RULE IS A VIEW"
            5. Inference - explaining it…..
               1. At its simplest… it’s a way to represent knowledge
               2. He uses a decision table example
                  1. i.e. a dept that issues validation to an adult vs child
                  2. Dept issues ID cards
                  3. But that dept is probably not responsible for defining child vs adult rules
               3. Gives u decoupling of knowledge
               4. UNDER 18 … MEANS THEY ARE A CHILD
               5. OVER 18 MEANS THEY ARE AN ADULT
               6. ITS ABOUT ENCAPSULATING KNOWLEDGE
            6. Look to java design patterns
            7. Can use inference to let different depts manage related data
            8. TO HIDE KNOWLEDGE THAT OTHER DEPTS DON’T NEED TO KNOW ABOUT
            9. LINKS
            10. from Mark Proctor to All Participants:
                http://blog.athico.com/2009/11/what-is-inference-and-how-does-it.html
            11. from Mark Proctor to All Participants:
                http://blog.athico.com/2010/01/drools-inference-and-truth-maintenance.html
            12. from Mark Proctor to All Participants:
                http://blog.athico.com/2010/02/jug-lille-videos-drools-expert-and.html
            13. from Mark Proctor to All Participants:
                http://en.wikipedia.org/wiki/Materialized_view

3. JCO -
   1. Rules are not just IF THEN
      1. Also can have probability factors
   2. Inference
      1. Forward
         1. Data driven
         2. Ideal where u already have all the data
         3. Used for
            1. Financial
            2. insurance
      2. Backward
         1. Goal driven
         2. Used for
            1. configuration problems
               1. Bell helicopters uses backward chaining for configuration
            2. Diagnostics
            3. scheduling

3. James says any system that does forward can also do backward , and vice versa ?

4. Pattern matching?

5. MONOTONIC REASONING?
   1. GO thru the system once, look at all IF THEN ELSE once

6. NON-MONOTONIC REASONING?
   1. Go thru the rules multiple times
   2. + my rulebase can grow in terms of size, i.e. adds rules to working memory

3. RH's thoughts on chaining.
   1. Some new slides I'm working on to explain the concepts to others

4. Recommends book on
   1. MYCIN
   2. Riley's book on ES

5. All
   1. JCO - Mgrs should never have to learn what backward and forward chaining is
      1. Or is it written in Cobol, assembler, Java, etc
      2. All they care about and need to know is the key P&L stuff
   2. RH & JT - disagree, mgrs need to understand at a high level what the engine does and how it adds value vs. hard-coding… JT & RH exchanged emails on this topic afterwards… JT has some great analogies / stories about this topic

2. Wrap up
   1. MP - recommends rulesposse see http://javaposse.com/
      1. q/a format
      2. JM - also sounds like javaranch website
   2. ORF2010
      1. JM is working on getting the website up shortly
      2. IF YOU WANT TO ASSIST VOLUNTEER, CALL JASON

3. Ideas on this monthly call
   1. 1 hr is about right
   2. Lets email questions ahead of time to the group
   3. LOOK INTO RECORDING THE WEBEX/Skype call

6. Website/membership update, BRxG AA Agility Alliance

7. WORKING GROUPS
   • Challenging business problems for the DRG to solve using BREs at their UG meeting
     • Hospitality/Travel, Car, Bank, Mortgage, EA
   • BR frameworks, templates, models,
     • Inference, Decision trees, Decision tables, Graphs
   • Logical rulebase models, designs, standards
   • Case studies to write & publish
   • Projects & Peer Reviews
   • Best practices of KA KR KE (knowledge elicitation / knowledge acquisition / knowledge representation / knowledge engineering)

8. NEXT STEPS
   1. Link-in with everyone on the call

9. SUGGESTIONS FOR NEXT TIME
   1. The state of BRE BRMS RBMS
   2. Issues, state of the BRE art, problems and solutions, tips & ideas, suggestions for next meeting/future calls
   3. News, blogs, projects, 5" presentations?
   4. Any other interesting topics?
   5. What do we want to actually DO, or ACCOMPLISH in the rules user grp

10. TRANSCRIPT

from Jason Morris to All Participants:
Joel: look at your volume control and see that your mic isn't muted
from Rolando Hernandez to All Participants:
ok i see you guys on the chat now
from Rolando Hernandez to All Participants:
lets get going
from Jason Morris to All Participants:
the night is young
from Mark Proctor to All Participants:
happy to mention something on inference
from Mark Proctor to All Participants:
rolando, want another take on that?
from Mark Proctor to All Participants:
http://blog.athico.com/2009/11/what-is-inference-and-how-does-it.html
from Mark Proctor to All Participants:
http://blog.athico.com/2010/01/drools-inference-and-truth-maintenance.html
from Mark Proctor to All Participants:
http://blog.athico.com/2010/02/jug-lille-videos-drools-expert-and.html
from Mark Proctor to All Participants:
http://en.wikipedia.org/wiki/Materialized_view
 
from Mark Proctor to All Participants:
The Fosdem representation in my blog links is probably a good one. it contains 3 examples:
Stateless
stateful execution,
inference and tms... in ways most people can understand.
 
from Jason Morris to All Participants:
Rolando: I think I there are better ways to illustrate these concepts :-)
from Mark Proctor to All Participants:
yeah, once drools does backward chaining, I'll have to borrow those images ;)
from Mark Proctor to All Participants:
I like the first two
from Mark Proctor to All Participants:
yeah, abstract though.
from Mark Proctor to All Participants:
oh that reminds me. there is another drools book "cook book" by Packt planned for Sept.
from Mark Proctor to All Participants:
looking for authors, even if only able to do a chapter. no money in it, but good way to spread the word. The two Drools books selling ok, about 1000 each so far.
from Jason Morris to All Participants:
what is old is new.... AI Spring is alive.... much untapped potential in practical AI apps
from Mark Proctor to All Participants:
http://javaposse.com/
from Jason Morris to All Participants:
http://www.javaranch.com