Professur Wissenschaftliches Rechnen
Wissenschaftliches Rechnen

Winter Semester 2019/20

Matrix Methods in Data Science (4V + 2Ü)

Lecture: Martin Stoll

Exercise: Michael Schmischke

News

  • The lectures and exercise groups will be held in English.
  • The fist lecture will be on Monday, 14th of October, 5. LE (15:30 - 17:00) in room 2/N105.
  • On Tuesday, 15th of October, 4. LE (13:45 - 15:15) in room 2/39/738, will be the first exercise.
  • Questions can be directed to Michael Schmischke.
  • On Tuesday, 22nd of October, there will be a lecture instead of an exercise. The room will be announced soon.
  • On Wednesday, 23rd of October, there will be NO lecture.
  • Group 2 will move their exercise from Tuesday to Monday in room 138 RH39/41.
  • The lecture will take place on Tuesday 13.30 in W0059.
  • On Wednesday, 13rd of November, there will be NO lecture.
  • For a nice python example on the PCA see here.
  • Mark Embree (Virgnia Tech) talking about the DEIM CUR.
  • Please register for your group in OPAL.
  • Gil Strang (MIT) talking about randomization in NLA here.
  • For the exam you will be handed a sheet with short quesitons and a programming task for which you will have 30 minutes to prepare. We provide a laptop with Python. Please arrive 30 minutes before your exam. Additional questions in the exam could look like the following here.

Content of the course

Motivating examples, Matrix factorisations for Classification and Learning: QR, SVD, Randomization, Nonnegative Matrix Factorisation, Numerical Tensor Methods, ...

Dates

Lectures
  • Monday, 5. LE (15:30 - 17:00), room 2/N105
  • Wednesday, 6. LE (17:15 - 18:45), room 2/B202
Exercise
  • Group 1: Tuesday, 3. LE (11:30 - 13:00), room 2/39/738
  • Group 2: Tuesday, 4. LE (13:45 - 15:15), room 2/39/738
Add the course to your own schedule via

Keine Lehrveranstaltung gefunden.

Exercise

  • The exercise sheets will be Jupyter notebooks. We will work on them together during the exercise time.
  • You can use the pool computers or your own devices. The computations will be handled by the faculty Jupyter Hub.
  • You do not have to use the Jupyter Hub, but we cannot provide any assistance if you decide to run Jupyter and Conda locally.
  • Every exercise will have (at least) a mandatory part which may be relevant for the exam. In some cases there will be an optional part that you can work on in your spare time.
  • The sheets will be available in OPAL.

Exam

  • The exam will consist of a 30-minute oral exam with preparation time beforehand.

Modul Affiliation

  • Matrix Methods in Data Science (M26).
  • "Forschungsmodul Numerische Mathematik (groß)" (FN3), "Forschungsmodul Optimierung (groß)" (F03) and "Forschungsmodul Angewandte Mathematik (groß)" (FM3).

Literature

  • Eldén, Lars; Matrix methods in data mining and pattern recognition
  • Kolda, Tamara G; Bader, Brett W; Tensor decompositions and applications SIAM review
  • Von Luxburg, Ulrike; A tutorial on spectral clustering Statistics and computing
  • Sorensen, Danny C; Embree, Mark; A deim induced cur factorization SIAM Journal on Scientific Computing
  • Higham, Catherine F; Higham, Desmond J; Deep Learning: An Introduction for Applied Mathematicians arXiv preprint arXiv:1801.05894
  • Gillis, Nicolas; The why and how of nonnegative matrix factorization Regularization, Optimization, Kernels, and Support Vector Machines