The Recommender
Challenge Hackathon
plista GmbH
2013/08/02
Torben Brodt
What is plista
● recommendation
● advertising
● network
● many big publishers in DE, AT, CH, ..
● "other articles you might be interested.."
● >8 billion impressions, clicks, engages, .. pM
Architecture
Tracking Success
● each time a recommender is chosen, plista
will track its success.. for context and
context combinations
???
Tracking Success
● "online evaluation" technology
● better than classical offline evaluation known
from papers?
● cooperation with TU Berlin, aided by state
???
The hackathon
● we open the data, you provide the
knowledge
● develop a recommender which implements
the http + json api
● plista will track the success, if you are smart,
be the winner for the the best recommender
● best is live, best is scalable and best will
work in industry
The hackathon
● many interesting people
● get to know developers using
○ PHP, Java, NodeJS, Python
○ Redis, Storm, Elastic Search
○ Apache Mahout, Lucene
○ ...
The hackathon
● http://contest.plista.com/
○ started 2 year ago
● New API in august
● News Recommender Challenge
○ ACM RecSys HongKong http://recsys.acm.org/
How to start (1/3)
register at
contest.plista.com
How to start (2/3)
● start implementation using examples
● http://contest.plista.com/wiki/example
How to start (2/3)
● start implementation using examples
● http://contest.plista.com/wiki/example
● have a github account?
● "fork" one of the example projects
● work on your local "clone"
● upload to your server
● enter url in your contest account
How to start (3/3)
● need a virtual server? ask us
● need old data? start replay from
webinterface
● try sending debug events from webinterface
● wait for team activation
● plista starts sending you real data
● your responses are displayed on real
publishers
Recommender ideas
● concentrate on implicit feedback
● think streaming / incremental
○ better to scale
○ faster results, new articles are better than old
articles?
● think about cross domain
○ contest is not allowed to mix items from different
domains/publishers
○ want knowledge of the full data, but candidate items
of a slice
How to go on?
improve the algorithms
● there will be a new api
● there will be more competition (SIGIR,
RecSys)
join the meetups: http://recommenders.de/
join the team: http://www.plista.com/career

#TOA13 - Tech Opoen Air Recommender Hackathon

  • 1.
    The Recommender Challenge Hackathon plistaGmbH 2013/08/02 Torben Brodt
  • 2.
    What is plista ●recommendation ● advertising ● network ● many big publishers in DE, AT, CH, .. ● "other articles you might be interested.." ● >8 billion impressions, clicks, engages, .. pM
  • 3.
  • 4.
    Tracking Success ● eachtime a recommender is chosen, plista will track its success.. for context and context combinations ???
  • 5.
    Tracking Success ● "onlineevaluation" technology ● better than classical offline evaluation known from papers? ● cooperation with TU Berlin, aided by state ???
  • 6.
    The hackathon ● weopen the data, you provide the knowledge ● develop a recommender which implements the http + json api ● plista will track the success, if you are smart, be the winner for the the best recommender ● best is live, best is scalable and best will work in industry
  • 7.
    The hackathon ● manyinteresting people ● get to know developers using ○ PHP, Java, NodeJS, Python ○ Redis, Storm, Elastic Search ○ Apache Mahout, Lucene ○ ...
  • 8.
    The hackathon ● http://contest.plista.com/ ○started 2 year ago ● New API in august ● News Recommender Challenge ○ ACM RecSys HongKong http://recsys.acm.org/
  • 9.
    How to start(1/3) register at contest.plista.com
  • 10.
    How to start(2/3) ● start implementation using examples ● http://contest.plista.com/wiki/example
  • 11.
    How to start(2/3) ● start implementation using examples ● http://contest.plista.com/wiki/example ● have a github account? ● "fork" one of the example projects ● work on your local "clone" ● upload to your server ● enter url in your contest account
  • 12.
    How to start(3/3) ● need a virtual server? ask us ● need old data? start replay from webinterface ● try sending debug events from webinterface ● wait for team activation ● plista starts sending you real data ● your responses are displayed on real publishers
  • 13.
    Recommender ideas ● concentrateon implicit feedback ● think streaming / incremental ○ better to scale ○ faster results, new articles are better than old articles? ● think about cross domain ○ contest is not allowed to mix items from different domains/publishers ○ want knowledge of the full data, but candidate items of a slice
  • 14.
    How to goon? improve the algorithms ● there will be a new api ● there will be more competition (SIGIR, RecSys) join the meetups: http://recommenders.de/ join the team: http://www.plista.com/career