Optimization Methods for Machine Learning

Description

Optimization techniques are used in all kinds of machine learning problems. This course gives an overview of many concepts, techniques, and optimization algorithms in machine learning and statistical pattern recognition. We also touch theory behind these methods (e.g., optimality conditions and duality theory). We also discuss how to choose and to set up the right optimization methods for different machine learning applications.

Teacher: Laura Palagi

Assistant: Tommaso Colombo (office hours:  on email appointment)

Register to the Google group OMML_2017-18  for receiveing info about the course (scheduling, timetable, teaching material etc.). Registration is possibile starting on September 25 until to October 31,  2017.

Topics include:

  1. basics of learning theory (error functions; VC theory; margins).
  2. Supervised learning:
    • deep networks
    • support vector machines
  3. Use of standard software is discussed (WEKA, LIBSVM, R, TensorFlow)