Topic outline

  • General

    This course goes beyond the standard classification and regression settings of supervised machine learning, and introduces ways to predict structured data (e.g. strings, graphs..). Some topics during this course include structured SVM, kernel dependency estimation, structured prediction energy networks, and optimisation approaches for solving the inference problem. 

    The course spans two periods: the first part of the course consists of mandatory lectures, and the second the graded project work. There is no exam.

    The course is aimed at postgraduate level. Course assumes background knowledge in machine learning and statistics, notably knowledge equivalent to CS-E4710 - Machine Learning: Supervised Methods and its prerequisites. Contents of CS-E4830 - Kernel Methods in Machine Learning are also recommended, but not mandatory.

    Restricted to 15 participants.