NSCPHD1233 Online Learning Summer School

Volume 2014/2015
Content
  • Subject area: Online Learning

 

Online learning is a broad subfield of machine learning covering problems involving iterative predictions/actions by the learner and revelation of the outcomes. It includes problems like stock market prediction, ads placement on the Internet, medical treatments, robotics, and many more. It involves dealing with continuous streams of data, handling non-stationary environments, coping with adversarial opponents, and many other challenging practical problems. Online learning has changed the way data is processed and unveiled a new world of approaches to personalization, large-scale data processing, and many other challenges of our data-intensive era.

 

  • Scientific content

 

The course will consist of 5 days of lectures and exercises and cover the following topics in online learning:

Day 1: Online learning with full information feedback and relations to optimization, to be taught by Shai Shalev-Shwartz.

Day 2 & 3: Online learning with limited (bandit) feedback in stochastic and adversarial environments, to be taught by Nicolò Cesa-Bianchi and Peter Auer.

Day 4: Reinforcement learning, to be taught by Csaba Szepesvári and Peter Auer.

Day 5: "The space of online learning problems" - interpolations between various online learning settings, to be taught by Yevgeny Seldin.

 

  • Learning outcome

 

After participating in the course the students should gain the following competencies:

  • Knowledge of basic concepts in online and reinforcement learning
  • Ability to formulate real-life problems in the language of online or reinforcement learning
  • Knowledge of algorithms and evaluation measures for online learning problems
  • Knowledge of lower bounds on the performance of online learning algorithms
  • Ability to apply online learning algorithms to problems in their research

Knowledge about the tools for analysis and development of online learning algorithms

Lectures and exercises
  • Category
  • Hours
  • Lectures
  • 40
  • Preparation
  • 10
  • Project work
  • 15
  • Total
  • 65
Credit
2,5 ECTS
Type of assessment
Continuous assessment under invigilation