Bern Data Science Day 2021
The Bern Data Science Day brings together data scientists from the University of Bern, the Bern University Hospital, sitem-insel and the Psychiatric University Clinic (UPD) for a unique conference on applied Data Science (DS). It gathers scientists from different domains to network and exchange ideas on emerging trends and research results in data science. The stimulating program will focus on extended poster sessions in lobby mode for networking facilitation, the keynote talks framing the event. Everyone working with or on aspects of data processing is invited to submit abstracts related to the below topics. Abstracts are reviewed by the scientific committee for oral and poster presentation. Vote winning contributions will receive prizes from our sponsors. Posters are reviewed and invited for voluntary upload to the BDSD Zenodo repository with DOI for reference.
BDSD 2021 is history. Stay tuned for the 2022 edition !
Important dates and deadlines
|Notification to authors:
|April 15 and April 21
Where and How
The event is scheduled online due to the pandemic measures. There will be one plenary room and poster session rooms which participants can stroll. The presenters will be present there for exchange and conversations. During the award session an online voting by the scientific committee and all participants will decide on the prize winning contributions.
The Zoom meeting link will be emailed to registered participans shortly before the event (Thursday evening or Friday morning). For proper participation in poster sessions, a Zoom client (desktop or mobile) is required. Please make sure in advance that it's installed and up to date. There will be an award vote after the poster sessions, with up to 5 votes per person. Please note the poster IDs which you would like to vote for. A Google account is required to ensure fair voting.
Agenda April 23 (see separate menu tab for detailed programme)
|08:45 - 09:00
|09:00 - 10:00
|Plenary with selected oral presentations
|10:15 - 12:00
|Poster session 1
|12:00 - 13:00
|13:00 - 14:00
|Keynote - Prof. Dr. Tamara Broderick from MIT
|14:15 - 16:00
|Poster session 2
|16:00 - 17:00
|Awards and closing session
|17:00 - 19:00
Anyone with interest and experience in the topics of Data Science, Machine Learning or Artificial Intelligence through their research can attend. Participation in the event is free of charge, but advance registration is required. To register, please fill out the registration form.
Submission of abstracts
Students, employees and participants in continuing education programs of the University of Bern, the Bern University Hospital, sitem-insel and the Psychiatric University Clinic (UPD) are warmly invited to submit abstracts related to one of the Data Science topics listed below. Abstracts will be reviewed by the scientific committee and authors will be invited to present a poster (electronically on zoom) at the event. A few selected posters will also be offered short oral presentations in the plenary sessions. After the event the contributions can be uploaded to a Bern Data Science Day repository which provides a digital object identifyer (DOI ) for reference.
Data Science (DS) call topics:
- DS algorithms with a view towards Machine Learning and Artifical Intelligence
- Mathematical and statistical foundations of DS
- DS applications and challenges in Medicine, Natural Sciences, and Engineering
- Digital Humanities and Economics
- Data representation, formatting, storage, and transfer
- DS and high-performance computing, e.g. work using the UBELIX cluster
- DS programming languages and softwares
- Ethical and legal aspects of DS
- DS AOB (any other DS business)
Please submit your abstract via this link. You may use this template. If you prefer not to have any submitted material publically exposed, apart from the showing in the event, please also note this in the abstract.
Keynote at 13:00 CET
Abstract: We propose a method to assess the sensitivity of data analyses to the removal of a small fraction of the data set. Analyzing all possible data subsets of a certain size is computationally prohibitive, so we provide a finite-data metric to approximately compute the number (or fraction) of observations that has the greatest influence on a given result when dropped. We call our resulting metric the Approximate Maximum Influence Perturbation. Our approximation is automatically computable and works for common estimators --- including (but not limited to) OLS, IV, GMM, MLE, and variational Bayes. We provide explicit finite-sample error bounds on our approximation for linear and instrumental variables regressions. At minimal computational cost, our metric provides an exact finite-data lower bound on sensitivity for any estimator, so any non-robustness our metric finds is conclusive. We demonstrate that the Approximate Maximum Influence Perturbation is driven by the signal-to-noise ratio in the inference problem, is not reflected in standard errors, does not disappear asymptotically, and is not a product of misspecification. We focus on econometric analyses in our applications. Several empirical applications show that even 2-parameter linear regression analyses of randomized trials can be highly sensitive. While we find some applications are robust, in others the sign of a treatment effect can be changed by dropping less than 1% of the sample even when standard errors are small.
- Event flyer can be found here.
Organization Committee and Contact
Sigve Haug, Alexander Kashev, Mauricio Reyes, Rahel Luder (email for questions), Kinga Sipos, Raphael Sznitman.
Prof. Dr. Dr. C. Beisbart, Prof. Dr. P. Favaro, Prof. Dr. D. Ginsbourger, PD Dr. S. Haug, Prof. Dr. M. Reyes, Prof. Dr. R. Sznitman, Prof. Dr. C. Tretter, Prof. Dr. A. Tzovara, Prof. Dr. J. Ziegel.