Module 3
Teachers
Examiner: Hossein Azizpour (azizpour@kth.se)
Teaching Assistants: Erik Englesson (engless@kth.se), Yiping Xie (yipingx@kth.se), Oskar Kviman (okviman@kth.se)
Aim
In the last module, we plan to familiarize students with some of the advanced topics of deep learning research including probabilistic and generative deep learning.
Schedule
Preparations: performed before June 14
Online Sessions: June 14-15
Assignments: performed between June 15-29 (or by August 15)
Assignments
Materials
Preparation: the preparation constitutes a quick recap of Bayesian modelling and a follow-up pre-study of variational inference. The notes can be found here: Module_3_Prestudy_VI.pdf Ladda ner Module_3_Prestudy_VI.pdf
Implementation Practicals: the list of implementation practicals to be done after the course sessions can be found here: Module 3 - List of Implementation Practicals
Deadlines
Home assignment. The standard deadline for the home assignment is June 29, but, in light of the covid-19 restrictions and to accommodate flexible planning of summer leaves we consider a second deadline of August 15. The second deadline will not incur any late submission penalty. The grading will be done by a month after the corresponding submission deadline.
Place (online)
All sessions are held online and using Zoom: https://kth-se.zoom.us/j/68686048311?pwd=K1g1Y2RqRVErU1lidWthaGRxRXAyQT09 Links to an external site.
Schedule
- June 14, 8:30am-10:30am
- Session 1: Introduction to Discriminative Modelling, Uncertainty Estimation and Bayesian Modelling
- Session 2: Uncertainty Estimation with VI Deep Networks
- June 14, 12:30am-14:30am
- Sessions 3-4: Examples of Simple and Effective Methods for Uncertainty Estimation with Deep Networks
- June 15, 8:30am-10:30am
- Session 5: Generative Models
- Session 6: Variational Auto Encoders
- June 15, 12:30am-14:30am
- Session 7: Auto-Regressive Deep Generative Models
- Session 8: Normalizing-Flow Deep Generative Models
- Offline: Generative Adversarial Networks