Autumn Academy 2022
We proudly present the first edition of the next regular online event in the annual calendar of MIDL: The Autumn Academy! Feel warmly welcomed to grab a cup of tea, curl up in a cozy blanket in front of your device and join a tutorial to train your research skills. And for this first edition, we chose one of the most important yet one of the most under-represented research skills to train: Reproducibility!
The Autumn Academy took place on November 30 2022 at 15:00 CET.
Tutorial: How to make your research reproducible?
Concerns about the reproducibility of deep learning research are more prominent than ever, with no clear solution in sight. Providing public source code is a powerful solution for one aspect of reproducibility, however the quality of these repositories vary widely between projects and might cause more harm than good. Evaluating all full paper submissions at MIDL shows that public repositories are becoming increasingly popular, however some tiny design flaws hinder their repeatability and usefulness over time. Based on these results and some commonly encountered problems, MIDL proposed a reproducibility guideline to help authors design a repository for their submission. (Following the guidelines is completely optional and their aim is not to regulate authors, but to help them.) During the tutorial we will discuss source code repositories and the proposed guideline, and through a series of discussion points we will have the chance to make changes to the guideline helping ourselves and future authors. No coding skills are required for the tutorial.
The tutorial will be led by Attila Simko from Umeå University.
Links to join the Sessions
We gave our best to select the optimal tools for each part of the event. The main program, i.e. the Tutorial, will be hosted in Zoom. You can join using the following link:
Following the main program, there will be a relaxed Get-Together starting from 18:15 CET. For that, we will switch to our good old MIDL event center in Gather.Town. You can join here:
Gather.Town is a great tool to meet and interact in cyberspace. If you feel overwhelmed or lost at first, don’t worry: You can find a short tutorial here.
Background: Reproducibility in MIDL
From its very beginnings in 2018, MIDL has always been committed to open science and transparency. The reproducibility of research, especially the reproducibility of code, models and implemented experiments, is a crucial part of that mission. However, there are typically few (or only very hidden) incentives to take that extra effort on your paper, and until now, MIDL has not been a big exception in that regard. Recently, a highly motivated team from Umeå University in Sweden provided evidence that only 22% of all MIDL papers are really reproducible. Luckily, they already proposed possible guidelines on reproducibility of MIDL papers, and we are now making these an integral part of our strategy and give our best to spread the word in the community. We firmly believe that the majority of not reproducible papers is not due to an active decision but to a lack of awareness. And what is the solution for that? An Autumn Academy solely dedicated to Reproducibility, open to the whole community!
Fact is: Reproducibility will become more and more important in the next years and will become a part of the review process in future MIDL editions. But besides that, your paper will benefit a lot since more people can use your methods more easily and thus your work will have a much higher impact. So, there are enough incentives to already join the movement today. And the easiest step is to join our Autumn Academy!
All times are Central European Time (CET):
|Check-In and Coffee Round
|14:50 - 15:00
|Welcome for Academy Participants
|15:00 - 15:15
|Part 1: Presentation and Hands-on Session
|15:15 - 16:30
|16:30 - 16:45
|Part 2: Hands-on Session
|16:45 - 18:00
|18:00 - 18:10
|Get Together in Gather.Town
- Alessa Hering, Radboudumc Nijmegen & Fraunhofer MEVIS, firstname.lastname@example.org
- Jannis Hagenah, Department of Engineering Science, University of Oxford, email@example.com