Difference between revisions of "Ne present 01"

From IPRE Wiki
Jump to: navigation, search
 
 
Line 1: Line 1:
 
'''Introduction to Recommender Systems'''
 
'''Introduction to Recommender Systems'''
What is an RS?
+
*What is an RS?
[What are different types of recommender systems]
+
*[What are different types of recommender systems]
 +
 
  
 
'''My Project'''
 
'''My Project'''
 +
 
''Part 1'': Exploring Recommendation Systems
 
''Part 1'': Exploring Recommendation Systems
look at different recommender systems
+
*look at different recommender systems
*by dif companies
+
**by dif companies
*dif methods
+
**dif methods
historical vantage-point: start w/ oldest & come to newest
+
*historical vantage-point: start w/ oldest & come to newest
what makes a good RS?
+
*what makes a good RS?
  
other directions:
+
*other directions:
*AI/machine learning used in RS
+
**AI/machine learning used in RS
*security/privacy issues?
+
**security/privacy issues?
*multi-agent systems, computer models for trust?
+
**multi-agent systems, computer models for trust?
  
 
''Part 2'': Designing a Recommendation System?
 
''Part 2'': Designing a Recommendation System?
plan is to design new RS -- use novel topic (e.g. Bryn Mawr course recommendations) or technique (hybrid of existing techniques?)
+
*plan is to design new RS -- use novel topic (e.g. Bryn Mawr course recommendations) or technique (hybrid of existing techniques?)
may deem infeasible or may prefer to explore RS research further
+
*may deem infeasible or may prefer to explore RS research further
in that case, will present analysis of current state of affairs re: recommendation systems & suggestions for further improvements
+
*in that case, will present analysis of current state of affairs re: recommendation systems & suggestions for further improvements
  
  
 
'''What I’ve Done So Far'''
 
'''What I’ve Done So Far'''
began reading articles re: RS
+
*began reading articles re: RS
*lay articles including Jeffrey O'Brien's [http://money.cnn.com/magazines/fortune/fortune_archive/2006/11/27/8394347/ ''The race to create a 'smart' Google''] -- '''it's interesting; check it out'''!!
+
**lay articles including Jeffrey O'Brien's [http://money.cnn.com/magazines/fortune/fortune_archive/2006/11/27/8394347/ ''The race to create a 'smart' Google''] -- '''it's interesting; check it out'''!!
*began mathy article Gediminas Adomavicius & Alexander Tuzhilin's [http://www2.computer.org/portal/web/csdl/abs/html/trans/tk/2005/06/k0734.htm ''Toward the Next Generation of Recommender Systems: A Survey of the State-of-the-Art and Possible Extensions'']
+
**began mathy article Gediminas Adomavicius & Alexander Tuzhilin's [http://www2.computer.org/portal/web/csdl/abs/html/trans/tk/2005/06/k0734.htm ''Toward the Next Generation of Recommender Systems: A Survey of the State-of-the-Art and Possible Extensions'']
 
*set up Wiki page (notes, sources, thesis description)
 
*set up Wiki page (notes, sources, thesis description)
 +
  
 
'''Next Week'''
 
'''Next Week'''

Latest revision as of 18:08, 30 January 2009

Introduction to Recommender Systems

  • What is an RS?
  • [What are different types of recommender systems]


My Project

Part 1: Exploring Recommendation Systems

  • look at different recommender systems
    • by dif companies
    • dif methods
  • historical vantage-point: start w/ oldest & come to newest
  • what makes a good RS?
  • other directions:
    • AI/machine learning used in RS
    • security/privacy issues?
    • multi-agent systems, computer models for trust?

Part 2: Designing a Recommendation System?

  • plan is to design new RS -- use novel topic (e.g. Bryn Mawr course recommendations) or technique (hybrid of existing techniques?)
  • may deem infeasible or may prefer to explore RS research further
  • in that case, will present analysis of current state of affairs re: recommendation systems & suggestions for further improvements


What I’ve Done So Far


Next Week

  • continue reading (& adding to wiki page)
  • specifically, look into old/original rec systems
  • want to know more about each of the three types of RS (personalized, social, item-based)
    • more details on how they work
  • look at companies -- how do they explain what they're doing (be wary of "we're so amazing"- type spin)
  • any info on classifying/rating rec systems -- how do you know if one is "better" than another?)