Machine Learning Consulting at CASUS

Friday 15 Jul 2022, 11:00 → 16:00 Europe/Berlin

CASUS office Görlitz

Steve Schmerler (HZDR)

As the pandemic regulations are being removed, we'd like to seize the chance and visit you as Helmholtz AI consultants. With this meeting, we like to learn

• what Machine Learning methods are being used,
• what your Machine Learning needs are and
• where we (as Helmholtz AI consultants) can support in the future

For this, we organize a "full-day" walk-in session.

What is this?

In order to connect with the wider research community in HZDR, we'd like to prototype a walk-in style service on a regular basis at our partnering Helmholtz centers. For this, we as consultants come by a Helmholtz center for one day. On this day, we will be available for focused discussions on ongoing or planned ML projects. We will offer seats for these sessions prior to this event.

The day is structured in two parts:

1. a general recap of what Helmholtz AI consultants have done in the last 2 years
2. the walk-in service session(s)

Each walk-in slot is 45 minutes in duration to allow for small breaks in between. If you like to talk to us during the walk-in sessions, please register here. Should we receive no registrations, we will adapt the timetable accordingly and be available in the time indicated.

Note: The walk-in slots will be assigned on a first-come-first-serve basis. Should the registration be closed as we run out of slots, please reach out to us to discuss your needs.

Who is this for?

Any Helmholtz staff is eligible to participate: from master/PhD student to Postdoc, staff scientist, non-science staff, institute lead or team lead. Our main expertise is in matter research but we also welcome scientists from other fields. The most important thing however will be that you bring data! If we don't have capacity to support you beyond the walk-in service, other teams in Helmholtz AI Consulting may have these resources.

The Helmholtz AI team at HZDR has expertise in ML problems applied to (mostly) matter research, including in particular: inverse problems, generative modeling, explainability, uncertainty quantification, time series analysis, anomaly detection, Gaussian process and kernel methods, image classification/ segmentation/ regression/ (un)supervised denoising, medical imaging, physics-informed neural networks.

What do people need to bring?

We like to ask participants of the walk-in consulting to prepare for the discussion by:

• preparing a presentation on their ideas or ongoing efforts in a structured way
  (see [1] for questions that should be answered) or
• bringing a data set and its visualization along or
• circulating the aforementioned material at least 1 week before the event.

The most important thing however will be that you bring data!

Registration will close end of Wed, Jul 13.

1 Two Years of Helmholtz AI Consulting for Matter

High level overview on the projects we did in the last 2-3 years. In-depth knowledge of machine learning is not required, but certainly helpful.

The talk will last about 20 minutes. The remainder of the time is aimed at an open Q&A session, where scientists can also express concerns, wishes or feedback.

15-casus-consultants-intro-Steinbach.pdf

12:00 Lunch

Session 1: Please Register for this.

In this block, we invite individuals or small teams to showcase their existing work using ML for science. We primarily encourage scientists or staff to bring unsolved problems where a "large" datasets exist. Please register for this!

2 Slot 1.1

3 Slot 1.2

4 Slot 1.3