Deep Learning, Advanced Course

Log in to your course web

You are not logged in KTH, so we cannot customize the content.

Welcome to Advanced Topics of Deep Learning!

*** Note 1. The course will run through both P1 and P2 for 2021 offering. P1 will include lectures, writing assignments, and project proposal (3HP) and P2 will have the students conduct their final project (3HP) ***

*** Note 2. Due to the situation caused by the spread of covid-19 the offering of the course in 2021 P1 is likely to be online. ***

*** Note 3. The course has limited seats (100-120 students). The applicants will be admitted according to the following priority groups ***

  1. Applicants who
    1. [have passed Deep Learning course(s) (DD2424 or DD2437) at KTH, and
    2. have passed other Machine Learning courses (DD2421 or DD2434 or similar) at KTH, and
    3. are from the following master programs: TMAIM, TSCRM, TCSCM Data Science Track] or
    4. [a PhD student whose thesis topic directly develops method for deep learning and
    5. have passed a basic deep learning course.]
  2. Applicants who
    1. [have passed Deep Learning courses (DD2424 or DD2437) at KTH, and
    2. have passed Machine Learning courses (DD2421 or DD2434 or similar) at KTH] or
    3. [a PhD student whose thesis topic mainly uses deep learning methods for applications and
    4. have passed a basic deep learning course].
  3. Applicants who
    1. [have passed Deep Learning courses (DD2424 or DD2437) at KTH,
    2. are from the following master programs: TMAIM, TSCRM, TCSCM Data Science Track] or
    3. [a PhD student whose thesis topic partially uses deep learning methods for applications and
    4. have passed a basic deep learning course].
  4. Applicants who
    1. have passed Deep Learning course(s) (DD2424 or DD2437) at KTH
  5. Any other applicant who have passed basic Deep Learning course(s) (prerequisite)

Applicants are admitted according to the priority group they fall into. If the applicants of the a group surpasses the number of available seats, the applicants will be sorted by their grade in the DL/ML courses they have completed  and the top applicants are admitted accordingly.

==============

As the topic suggests the course will delve into some advanced topics of deep learning. As such it requires solid background in core knowledge of deep networks (e.g. DD2424), machine learning in general, and math (calculus, probability theory and algebra). During the course we will go through recent advances in the field of deep learning which essentially means seminal papers from recent (~2014-2019) top venues such as ICML/NIPS/ICLR/JMLR/CVPR/ECCV/ICCV/EMNLP/ACL. This would mean that not having the proper background can make understanding and passing the course quite hard. The advanced topics we will cover are still tentative but includes probabilistic deep networks, uncertainty estimation, deep generative models, meta-learning, understanding deep networks, and the like.

The course will also involve reading, commenting and possibly presenting papers from the recent publications as individual assignments. The final form of assessment will be either writing a full review essay on the advanced topics or a an implementation project which more likely will be done in groups. The tentativeness of the assignment type and final evaluation is due to the unknown number of students who will register for this first round of the course and will be announced in the first lecture in August.

Teachers

Feedback News