Syllabus

1. Introduction
2. Sampling & Streaming
   • Motivations
   • Techniques for sampling large datasets (e.g., Poisson & Bernoulli sampling, Reservoir sampling)
   • Learning from continuous data-streams
     • Linear models (e.g., Naive Bayes, Perceptrons)
     • Non-linear models (e.g., Very fast decisions trees, SVMs, Neural Nets)
3. Hashing
   • Hash indexes
   • Hashing for exact matches (e.g., equi-join)
   • Hashing for similarity detection (e.g., LSH)
   • Hashing to handle an unknown number of features (e.g., count-min sketch) (If time permits)
4. Parallel & distributed computing
   • Computing architectures: shared memory architectures, clusters, grids
   • Parallel computing paradigms: multi-threading, message passing, MapReduce
   • Performance evaluation
   • Which configuration for which problem?
5. Hands-on with Hadoop
   • Analysis of GPS tracking data with Hadoop
   • Sampling versus distributed computing
   • Every participant gets an Hadoop cluster!
6. Discussion

Disclaimer: the syllabus is subject to change without notice.

The distributed computing infrastructure for the hands-on exercises with Hadoop is provided by the Flemish Supercomputer Centre.