A vacancy for a postdoc is available at our department (Data Analysis and Mathematical Modelling) of Ghent University.
Scientific research is broadly defined and can be about data analysis or mathematical modelling for natural and biological processes applied in one or more research areas at our faculty of bioscience engineering.
One possible topic includes experimental design problems within the field of machine learning (e.g. active learning, optimum designs for neural networks, ...)
There is a 30% teaching duty in Dutch associated to the position.
For more info and the application procedure, click here<https://career012.successfactors.eu/career?company=C0000956575P&career_job_…>
Deadline for application is 24/09
[cid:image001.png@01D75C87.92473700]<https://www.ugent.be/bw>
Prof. dr. Stijn Luca
Assistant professor
T +32 9 264 59 34
Department Data Analysis and Mathematical Modelling
Research Unit BIOSTAT: Biostatistics
www.biostat.ugent.be<http://www.biostat.ugent.be/>
Campus Coupure, Block A 1st floor 110.068, Coupure links 653, B-9000 Ghent, Belgium directions<https://www.google.com/maps/dir/Coupure+Links+653,+9000+Gent/@51.0527795,3.…>
T administration office +32 9 264 59 32
www.ugent.be<https://www.ugent.be/bw/en>
e-maildisclaimer<https://helpdesk.ugent.be/e-maildisclaimer.php>
Dear all,
It is my pleasure to announce the following CWI Machine Learning seminar.
Speaker: Johanna Ziegel (University of Bern)
Title: Valid sequential inference on probability forecast performance
Date: Friday 10 September, 14:00
Location: CWI L016 and on Zoom (link will follow)
This seminar will be held live at reduced capacity, and will be
live-streamed on zoom for remote attendance.
Please find the abstract below.
Hope to see you then.
Best wishes,
Wouter
Details:
https://portals.project.cwi.nl/ml-reading-group/events/valid-sequential-inf…
============
Valid sequential inference on probability forecast performance
Johanna Ziegel (University of Bern)
Probability forecasts for binary events play a central role in many
applications. Their quality is commonly assessed with proper scoring
rules, which assign forecasts a numerical score such that a correct
forecast achieves a minimal expected score. In this paper, we construct
e-values for testing the statistical significance of score differences
of competing forecasts in sequential settings. E-values have been
proposed as an alternative to p-values for hypothesis testing, and they
can easily be transformed into conservative p-values by taking the
multiplicative inverse. The e-values proposed in this article are valid
in finite samples without any assumptions on the data generating
processes. They also allow optional stopping, so a forecast user may
decide to interrupt evaluation taking into account the available data at
any time and still draw statistically valid inference, which is
generally not true for classical p-value based tests. In a case study on
postprocessing of precipitation forecasts, state-of-the-art forecasts
dominance tests and e-values lead to the same conclusions.
Dear All,
A video abstract can help the reader to have a clear understanding of
the paper. We are pleased to announce that the Robotics 2021–2022 Best
Video Abstract Award is open for applications. Authors are encouraged to
provide a video abstract when submitting their manuscripts in order to
improve the accessibility of papers. The winner will be selected through
a vote by our readers and Award Committee from the papers published in
Robotics.
Timeline:
– Deadline for video abstract submission: 31 December 2022;
– Voting period: 1 January 2023–31 January 2023;
– Date for announcement of the winner: 28 February 2023.
Prize:
There will be one awardee for this award, and the winner will receive
the following:
– CHF 500 (Swiss francs);
– A waiver to publish a paper in Robotics (submitted before the end of
2023);
– An electronic certificate.
Who can participate?
Authors who have written articles and reviews published in 2021–2022
with a video abstract.
For detail information, please click:
https://www.mdpi.com/journal/robotics/awards/1389
PLEASE NOTE: All video abstracts will be assessed for editorial
suitability and quality by the editorial team.
Please upload your video abstract when submitting your manuscript if you
are interested in the award. For the authors of published papers, please
send a video and a manuscript link to the Robotics Editorial Office
(robotics(a)mdpi.com) if you would like to apply for the award.
Charlene Dong
Managing Editor
Email: charlene.dong(a)mdpi.com
Robotics (http://www.mdpi.com/journal/robotics/)
10th Anniversary of Robotics: A series of special content and events at
https://www.mdpi.com/journal/robotics/anniversary
Dear colleagues,
Apologies for cross-posting. We have an opening for one 18 months postdoc position "Deep Learning for Human-Robot Interaction" at TU Delft, Netherlands.
This project will focus on various aspects of human robot interaction leveraging deep learning methods. Robots can learn to interact with humans but also from interactions with humans. In both cases understanding human behavior is crucial. Human state and intentions are very ambiguous and uncertain, fusing information from multiple sensing and input modalities might allow the robot to disambiguate. For example, in imitation learning approaches instructions from one modality typically are not complete. Here the challenge lies in learning which information, from which modality is relevant for the task, and which is not.
This vacancy is part of the project Open Deep Learning Toolkit for Robotics (OpenDR) https://opendr.eu/ .
For more information and instructions on how to apply (deadline September 15th), see
https://www.tudelft.nl/over-tu-delft/werken-bij-tu-delft/vacatures/details?…
Best regards,
Jens
--
Dr.-Ing. Jens Kober
Associate Professor, Cognitive Robotics department
Delft University of Technology
Mekelweg 2 (3mE Building, Office F-2-380)
2628 CD Delft, The Netherlands
Tel.: +31 (0)15 27 85150
http://www.jenskober.de
Forwarding on behalf of Joris Mooij:
We are seeking six PhD candidates and a postdoctoral researcher for the
recently established Mercury Machine Learning Lab (MMLL).
In this lab, researchers from the University of Amsterdam (UvA) and
Delft University of Technology (TU Delft) will be working together with
data scientists from Booking.com to develop the machine learning
foundations for a new generation of recommendation systems. Motivated by
real-world problems faced in industry, we will investigate fundamental
problems regarding generalization and bias removal in data analysis, and
develop more usable machine learning and reinforcement learning techniques.
As part of the MMLL initiative, the University of Amsterdam and the
Delft University of Technology are inviting applications for the
following six fully funded PhD positions:
Position 1: Bridging online and offline evaluation (supervisor:
Maarten de Rijke, UvA)
Position 2: Avoiding filter bubbles by correcting for selection
bias (supervisor: Joris Mooij, UvA)
Position 3: Domain generalization and domain adaptation
(supervisor: Joris Mooij, UvA)
Position 4: Training NLP Models for Better Generalisation
(supervisors: Ivan Titov & Wilker Aziz, UvA)
Position 5: Model-based Exploration (supervisors: Matthijs Spaan &
Frans Oliehoek, TU Delft)
Position 6: Parallel Model-based Reinforcement Learning
(supervisors: Matthijs Spaan & Frans Oliehoek, TU Delft)
More details on the contents of the MMLL research projects are provided
on the MMLL webpage, https://icai.ai/mercury-machine-learning-lab/
In addition, the University of Amsterdam invites applications for a
postdoctoral researcher to assist with the research and the supervision
of the PhD students.
For further information, see the official job advertisements at:
https://www.uva.nl/en/content/vacancies/2021/07/21-579-phd-4-and-postdoc-1-…https://www.tudelft.nl/over-tu-delft/werken-bij-tu-delft/vacatures/details?…
The application deadline is August 23rd, 2021.
-------------------------------------------------------------
Joris Mooij
Professor in Mathematical Statistics
University of Amsterdam
http://www.jorismooij.nl/
(apologies for cross-posting)
Do you have what it takes to push reinforcement learning beyond the
realm of games?
TU Delft (Netherlands) offers 2 PhD positions focusing on reinforcement
learning in the recently established Mercury Machine Learning Lab
(MMLL). In this lab, researchers from the University of Amsterdam and
Delft University of Technology will be working together with data
scientists from Booking.com to develop more usable reinforcement
learning and other machine learning techniques.
Reinforcement learning is a promising approach to learn to control
decision making problems that extend over time, but so far applications
have been largely limited to synthetic settings such as games. Motivated
by real-world problems faced in industry, we will investigate
fundamental problems in reinforcement learning. For instance, we will
study effective exploration in non-stationary environments and learning
using many parallel trials. The candidates will be jointly supervised by
Dr. M.T.J. Spaan and Dr. F.A. Oliehoek.
The MMLL collaboration provides the unique opportunity to test AI
techniques in the real world, allowing new machine learning methods to
be safely developed for wide application, for example in mobility,
energy or healthcare. In addition to the existing researchers, the
Mercury Machine Learning Lab will comprise six PhD candidates and two
postdocs who will work on six different projects related to bias and
generalisation problems over the course of the next five years.
Further details and an application form can be found via the following
links.
https://icai.ai/mercury-machine-learning-lab/https://www.tudelft.nl/over-tu-delft/werken-bij-tu-delft/vacatures/details?…
Application deadline for full consideration: 23 August 2021
Informal enquiries: Matthijs Spaan (m.t.j.spaan(a)tudelft.nl) and Frans
Oliehoek (f.a.oliehoek(a)tudelft.nl).
Dear all,
On Thursday the 22nd of July, at 3PM CET, we will have the last speaker in our seminar before the summer stop. Ilias Bilionis from the School of Mechanical Engineering at Purdue University will talk about physics-informed neural networks, high-dimensional uncertainty quantification, automated discovery of physical laws and complex planning, in the context of work done at NASA's Resilient Extra-Terrestrial Habitats Institute. You can find the zoom link and abstract below. Feel free to share the zoom link with anyone that might be interested.
Kind regards,
Wouter Edeling
22 Jul. 2021 15h00 CET: Ilias Bilionis (School of Mechanical Engineering, Purdue University): Situational awareness in extraterrestrial habitats: Open challenges, potential applications of physics-informed neural networks, and limitations
I will start with an overview of the research activities carried out by the Predictive Science Laboratory (PSL) at Purdue. In particular, I will use our work at the Resilient Extra-Terrestrial Habitats Institute (NASA) to motivate the need for physics-informed neural networks (PINNs) for high-dimensional uncertainty quantification (UQ), automated discovery of physical laws, and complex planning. The current state of these three problems ranges from manageable to challenging to open, respectively. The rest of the talk will focus on PINNs for high-dimensional UQ and, in particular, on stochastic PDEs. I will argue that for such problems, the squared integrated residual is not always the right choice. Using a stochastic elliptic PDE, I will derive a suitable variational loss function by extending the Dirichlet principle. This loss function exhibits (in the appropriate Hilbert space) a unique minimum that provably solves the desired stochastic PDE. Then, I will show how one can parameterize the solution using DNNs and construct a stochastic gradient descent algorithm that converges. Subsequently, I will present numerical evidence illustrating this approach's benefits to the squared integrated residual, and I will highlight its capabilities and limitations, including some of the remaining open problems.
Join Zoom Meeting
https://cwi-nl.zoom.us/j/85892880510?pwd=QS9vaUxDWHBjWVp1djdIaVZvSmhUZz09
Meeting ID: 858 9288 0510
Passcode: 480369
Dear all,
I'm forwarding on behalf of a contact at KLM the following interesting
position:
Utrecht University has an vacancy for an associate professor on
Algorithms, Optimization and/or AI in Aviation in collaboration with KLM.
https://www.uu.nl/en/organisation/working-at-utrecht-university/jobs/associ…
The application deadline is 15 August 2021.
Cheers,
Wouter
Dear all,
Before the summer really starts, we have a very interesting invited
speaker in the thematic seminar on Friday next week:
*Gergely Neu* (Universitat Pompeu Fabra, http://cs.bme.hu/~gergo/)
*
Friday June 25*, 16.00-17.00
Online on Zoom:
https://uva-live.zoom.us/j/81805477265
Meeting ID: 818 0547 7265
Please also join for online drinks after the talk.
*Information-Theoretic Generalization Bounds for Stochastic Gradient
Descent*
We study the generalization properties of the popular stochastic
gradient descent method for optimizing general non-convex loss
functions. Our main contribution is providing upper bounds on the
generalization error that depend on local statistics of the stochastic
gradients evaluated along the path of iterates calculated by SGD. The
key factors our bounds depend on are the variance of the gradients (with
respect to the data distribution) and the local smoothness of the
objective function along the SGD path, and the sensitivity of the loss
function to perturbations to the final output. Our key technical tool is
combining the information-theoretic generalization bounds previously
used for analyzing randomized variants of SGD with a perturbation
analysis of the iterates.
Seminar organizers:
Tim van Erven
Botond Szabo
--
Tim van Erven <tim(a)timvanerven.nl>
www.timvanerven.nl