Lecture: Numerical Algorithms for Visual Computing and Machine Learning

Course

  • Lecturer(s):
  • Start: 19.10.2021
  • Dates: Tue. 14 (c.t.) - 16, online
  • Course number: MA-INF 2317
  • Curriculum: Master
  • Effort: 4.0 SWS / 6 CP
  • Exams: 22.02.2022 10-12, 22.03.2022 10-12 (both dates are tentative)

Exercises

  • Tutor(s):
    • Dongliang Cao
  • Start: Wed. 27.10.2021
  • Dates: Wed. 12 (c.t.) - 14, online

Description

This module focuses on numerical methods that frequently occur in the fields visual computing (VC) and machine learning (ML). In addition to algorithms, this module will also cover modelling aspects that are relevant for solving practical problems in VC and ML. The contents include:

  • Error analysis and conditioning of problems
  • Linear systems (solvability, algorithms, stability, regularisation), and applications and modelling in VC and ML (e.g. linear regression, image alignment, deconvolution)
  • Spectral methods (eigenvalue decomposition, singular value decomposition, respective algorithms), and their applications and modelling in VC and ML (e.g. clustering, Procrustes analysis, point-cloud alignment, principal components analysis)
  • Numerical optimisation (gradient-based methods, second-order methods, large-scale optimisation) and applications and modelling in VC and ML.