Learning/Event Service
Minutes: 2 November 1998 6:30 Smith Hall 100
PRESENT
  • Rudy Setiawan
  • Jon Hsieh
  • James Lampe
  • Yun-Ching Lee
  • Wing Ling Leung
  • Jon Wildstrom
  • Andy Zimdars
  • Eric Stein
  • Qiang Rao (Authentication team liason)
STATUS
  • Rudy:
    • Leading the team through Object Design phase
  • Andy:
    • Architecture team has a partially completed SDD
  • Wing:
    • Preparing the SDD presentation on concurrency
    • All subsystems run concurrently, but has sequential procedures
  • James:
    • Volunteered for leader position during Implementation phase
  • Yun-Ching:
    • Volunteered for minute taker position (until the end of time?)
    • Preparing the SDD presentation on System Control
  • Jon H:
    • Needs to clarify DB trigger issues with Database team
  • Jon W:
    • All documents are in CVS
DISCUSSION
  • SDD issues
    • Learning system exists in PAID server only
    • SDD has Learning team performing software routing
      • Hardware solution already exists
  • Question from Authentication Liason
    • Need to track authentication protocol usage/availablity as user moves between different regions, need to store this information in user profile
    • Need to encrypt communication between user client and dealer server
    • Suggestion:  designated a globally compatible standard
  • Question from UI Liason
    • UI <-> Learning information passing
    • Discuss on BBoard
  • Object Design Issues
    • Event Service
      • Network has something similar to event services, attempt to minimize duplication of effort
      • Naming Service
        • Who defines name?  Architecutre?
        • Not many groups realize they need events
    • Learning
      • What is in DataMiner?
        • Generalize to have multiple contents/algorithms
        • Inputs and Outputs: fill out the chart
Input Output
car model, location
etc.		 destination: cache/database
etc.
        • Global vs Local representation of dealers:  use individual representation
        • Replication issues with PAID and STAR server:  how often is it in sync?  Let DataMiner rebuild BehaviorFile on other servers if one PAID server goes down
        • DataMiner picks the best algorithm
          • Decision Tree
          • Genetic Algorithm
          • Statistical?
          • Neural Net?
        • Use Rudy's Learning algorithm broker architecture for DataMiner
      • What is in BehaviorFile
        • Storage:  in database or as a serialized object on local filesystem?
        • Contents
          • Andy: generalize dealers over geographic area
          • Jon H. & Wing: Separate behavior per learning task
          • Resolution:  Separate behavior per learning task
      • Scheduler should spawn DataMiner as necessary, but with predefined maximum sleep time, and use STAR DB to replicate learning data
      • LogDB should be logged and stored in database
ACTION ITEMS
  • All:
    • Discuss UI Liason's question on BBoard
    • Confer with teams re: events and API issues (same assignments as concurrency research)
    • Study scenarios and fill in DataMiner input/output chart (see above)
    • Homework 3!
  • Eric:
    • Seek clarification on software routing
  • James:
    • attend Authentication meeting/BBoard to resolve protocol usage preference issue
  • Rudy:
    • Find out how much of event services is Netowrk Team doing
    • Post DataMiner's algorithm interface
  • Andy:
    • Find out from Architecture team on who defines event naming
  • Jon H.:
    • Find out what DB team provides in term of DB triggers
  • Wing:
    • Evaluate learning algorithms
CRITIQUE
  • We had too much to discuss but not enough time to do it all in meeting
  • I should post minutes earlier next time

This page is hosted by the Chair for Applied Software Engineering of the Technische Universität München.
Imprint (Impressum)