Niklas Karlsson, 9 Feb 2016: Difference between revisions

From Murray Wiki
Jump to navigationJump to search
No edit summary
No edit summary
Line 8: Line 8:
* 10:00 am: Yisong Yue, 303 Annenberg
* 10:00 am: Yisong Yue, 303 Annenberg
* 10:30 am: Adam W, 215 Annenberg
* 10:30 am: Adam W, 215 Annenberg
* 11:00 am: Joel Tropp, 307 Annenberg
* 11:00 am: Venkat Chandrasekaran, 300 Annenberg
* 11:45 am: seminar setup
* 11:45 am: seminar setup
* 12:00 pm: seminar, 105 ANB
* 12:00 pm: seminar, 105 ANB

Revision as of 04:27, 4 February 2016

Niklas Karlsson will be visiting Caltech on 9 Feb 2016. Please sign up below if you would like to meet with him, listing your name and office number.

Schedule

9 Feb 2016 (Tue)

  • 9:00 am: open
  • 9:30 am: open
  • 10:00 am: Yisong Yue, 303 Annenberg
  • 10:30 am: Adam W, 215 Annenberg
  • 11:00 am: Venkat Chandrasekaran, 300 Annenberg
  • 11:45 am: seminar setup
  • 12:00 pm: seminar, 105 ANB
  • 1 pm: lunch with Yisong (Chandler)
  • 1:45 pm: open
  • 2:30 pm: open
  • 3:15 pm: open
  • 4:00 pm: open
  • 4:45 pm: open
  • 5:30 pm: depart from Ath for BUR (Caltech car)

Seminar: 9 Feb (Tue), 12-1 pm, 105 ANB

TITLE: Modeling and Control of Online Advertising

ABSTRACT: Internet advertising is a huge and growing industry with many interesting and challenging engineering problems. This talk first reviews the evolution of optimization paradigms adopted for online advertising from 1998 to 2015. Thereafter we discuss methods for how to describe online advertising as a high-dimensional dynamical system, and how on top of this to apply conventional engineering principles for control and optimization. We show how this on-the-surface non-standard engineering problem on a proper abstraction level can be represented in terms that engineers are familiar with. In the process of developing a model that describes the dynamics, we highlight unique aspects of the system, making it challenging as well as inviting for further research. The challenges involve non-linearities, time-variability, randomness, uncertainties, latency, and coupling effects. We finally sketch on a solution for advertising campaign optimization, where the objective is to deliver an advertising budget smoothly throughout a campaign flight at the smallest possible cost to the advertiser.