Due to the restrictions on physical meetings that are imposed
during the spread of the Corona virus, the classes will be done
online using Zoom video conferencing and Moodle during the first
weeks of April.
See the link for the next Zoom online lecture in the schedule
below. NB! the Zoom links are different for every lecture and ubung.
If this is your first time using Zoom, please have a
look here
on how to install and join Zoom meetings.
The remaining schedule for online lectures will
tentatively follow the old plan, but there may be changes. For
further information on how and when to attend the Zoom lectures
and on course material, please register and look at the course's
Moodle page.
Bayesian inference; stochastic processes; the Fokker-Planck equation, discrete time, linear and nonlinear filtering algorithms, numerical methods for data assimilation.
Lectures and ubungs: Hakon Hoel, office: C364, Kackertstr. 9, email: hoel AT uq.rwth-aachen.de
Main literature: "Filtering and Prediction: A Primer" by Fristedt, B, Jain, N. and Krylov, N., 1st ed. AMS (2007), and "Data assimilation" by Law, K., Stuart, A., & Zygalakis, K., 1st ed. Springer (2015).
Supplementary:
Most weeks there will be 4 SWS lectures and 2 SWS ubungs (see plan below for exceptions).
Bonus points equivalent to 0.1*MAX_EXAM score is awarded to those giving a (roughly) 20 minutes presentation on an important topic/paper in data assimilation at specified times in early July.
Suggested topics and papers (text marked in red are taken topics):
Final oral exam 31/7 from 9AM to 5PM at Kackertstr. 9, room C301. Credits: 9 ECTS.
The plan will be updated during the semester.
Literature abbreviations: (Fristedt, Jain and Krylov) = FJK, (Law, Stuart and Zygalakis) = LSZ, (Sanz-Alsons, Stuart, Taeb) = SST, (S. Sarka) = Sar, (Reich and Cotter) = RC, (E, Li and Vanden-Eijnden) = ELV-E
Date | Time and place | Content |
---|---|---|
Mon Apr 6 | 10:30-12 Zoom online lecture. |
Lecture slides
Content: Overview of the course. Review of probability
theory for discrete random variables: (FJK Chapter 1.1-1.2,
supplementary Durrett 1.1-1.3)
|
Thu Apr 9 | 8:30-10 Zoom online lecture | Lecture slides
Content: discrete random variables, independence,
moments, distributions,
(FJK Chapter 1.2-1.3)
Annotated lecture slides
|
Thu Apr 16 | 8:30-10 Zoom online lecture | Lecture slides
Content: Conditional probability and expectation (FJK Chapter 1.3) Annotated lecture slides |
Fri Apr 17 |
10:30-12 Zoom online ubung,
|
Ubung exercises |
Mon Apr 20 | 10:30-12 Zoom online lecture | Lecture slides
Content: Random walks and convergence of random variables (FJK Chapter 2.1)
Annotated lecture slides and randomWalk2d.m
|
Thu Apr 23 | 8:30-10 Zoom online lecture | Lecture slides Content: Discrete-time Markov chains (FJK Chapter 2.2-2.3) Annotated lecture slides |
Fri Apr 24 | 10:30-12 Zoom online ubung | Ubung 2 exercise sheet for ubung 2 |
Mon Apr 27 | 10:30-12 Zoom online lecture, |
Lecture slides Content: Discrete-time Markov chains, filtering and smoothing. (FJK 2.2-2.3,3.1,3.3,3.4) Annotated lecture slides
|
Thu Apr 30 |
8:30-10 Zoom online lecture, |
Lecture slides Content: Continuous state-space random variables, probability density functions and conditional pdfs and expectations. (Durrett chp 1, FJK chp 4). Annotated slides |
Mon May 4 | 10:30-12 Zoom online lecture | Lecture slides Content: Conditional expectations for random variables in general, projections onto subspaces of L^2(Omega), and Bayesian inverse problems and well-posedness (Durrett chp 1, LSZ chp 1 and overlap with SST chp1) Annotated slides |
Thu May 7 | 8:30-10 Zoom online lecture | Lecture slides Lecture: Stability of Bayesian inversion, TV and Hellinger distance. (SST chapter 1) Annotated |
Fri May 8 | 10:30-12 Zoom online ubung |
Ubung 3 exercise sheet for ubung 3
|
Mon May 11 | 10:30-12 Zoom online lecture, |
Lecture slides Content: Bayesian inversion in setting with exact observations but uncertain model. Bayesian inversion in linear-Gaussian setting, convergence of the posterior in the zero-noise limit. (SST chp 2) Annotated |
Thu May 14 | 8:30-10 Zoom online lecture, |
Lecture slides Content: Gaussian fitting of posterior densities, parameter optimization, Kullback-Leibler divergence. (SST 3, 4) Annotated |
Fri May 15 | 10:30-12 Zoom online ubung |
Ubung exercise sheet for ubung 4
|
Mon May 18 | 10:30-12 Zoom online lecture, |
Lecture slides
Content: Monte Carlo method and importance sampling of posterior densities.
Markov chain Monte Carlo method. Markov chains in discrete-time continuous
state space. (SST 5,6, LSZ 2)
Annotated
|
Fri May 22 | 10:30-12 Zoom online ubung |
Ubung exercise sheet for ubung 5 |
Mon 25 May | 10:30-12 Zoom online lecture, |
Lecture slides Content: Smoothing in discrete-time continuous state space. (LSZ chp 2, supplementary SST chp 7) |
Thu 28 May | 8:30-10 Zoom online lecture, |
Lecture slides Content: Filtering in discrete-time continuous state space. The linear-Gaussian settingr; the Kalman filter. (LSZ chp 2, and 4.1, supplementary SST chp 8) Annotated |
Fri 29 May | 10:30-12 Zoom online ubung |
Ubung exercise sheet for ubung 6 |
Thu Jun 4 | No lecture: lecture free Pentecost week at RWTH | |
Fri Jun 5 | No ubung: lecture free Pentecost week at RWTH |
|
Mon 08 Jun | 10:30-12 Zoom online lecture, |
Lecture slides [with some correction to the motivation of ExKF] Content: Variational approximate filtering methods: 3DVAR and 4DVAR (SST chp 9 and LSZ chp 4). Nonlinear filtering methods: the extended Kalman filter. (SST chp 10 and LSZ chp 4). |
Fri 12 Jun | 10:30-12 Zoom online ubung |
Ubung exercise sheet for ubung 7 |
Mon 15 Jun | 10:30-12 Zoom online lecture, |
Lecture slides Lecture: ExKF and EnKF, and convergence of nonlinear filtering methods(LSZ chp 4 and SST chp 10). |
Thu 18 Jun | 8:30-10 Zoom online lecture, |
Lecture slides Content: Particle filtering (LSZ chp 4, SST chp 11) |
Fri 19 Jun | 10:30-12 Zoom online ubung |
Ubung exercise sheet for ubung 8 |
Mon 22 Jun | 10:30-12 Zoom |
Time to work on student presentations. If you have questiosn, I will be present on Zoom. |
Thu 25 Jun | 8:30-10 Zoom online lecture, |
Lecture slides Content: Alternative dynamics for particle filters (SST chp 12), Stochastic processes, filtrations, Wiener processes and Ito integrals (ELV-E 5.1-5.3, 6.2-6.4, 7.1) |
Fri 26 Jun | 10:30-12 Zoom |
Time to work on student presentations. I will be present on Zoom if you would like to discuss your preparations. |
Mon 29 Jun | 10:30-12:00 Zoom online lecture, |
Lecture slides
Content: Ito integrals, Ito SDE, numerical integration of SDE,
and the Fokker-Planck equation (ELV-E 7.1-7.3, 7.5, 8.1).
|
Thu Jul 2 | 8:30-10 Zoom student presentations |
Student presentations |
Fri Jul 3 | 10:30-12 |
Cancelled: Student presentations moved to Monday |
Mon Jul 6 | 10:30-12 Zoom online presentations, |
Student presentations. |
Thu Jul 9 | 8:30-10 Zoom online lecture |
Lecture slides
Lecture: Data assimilation for stochastic differential equations, and model selection/parameter estimation.
(Material is mainly self-contained, but with some extracts from RC chp 4.4 and 9.1 on model selection/parameter estimation) |
Fri Jul 10 | 10:30-12:00 Zoom ubung |
Ubung exercise sheet for ubung 9 |
Mon Jul 13 | 10:30-12 Zoom online lecture, |
Lecture slides Lecture: Filtering for continuous-time dynamics and observations (LSZ chp 6.1.1, 6.2, 6.4, 8.1, 8.2, supplementary Oksendal chp 6) and filtering in high dimensions (RC 8.1-3). |
Thu Jul 16 | 8:30-10 Zoom online lecture |
Lecture: presentations by Dmitry Kabanov on
Ensemble
Kalman Inversion applied to machine learning, and by Luis Espath
on Bayesian optimal experimental design.
|
Fri Jul 17 | 10:30-12:00 Zoom course summary link: https://rwth.zoom.us/j/99456413043 meeting ID:994 5641 3043 |
Oral exam topics.Course summary and questions |