Zum Inhalt springen

Details

We are incredibly happy to announce our next Meetup on November 24th at Zeiss.
Format:

  • 3 talks (ca. 20+20 and ca. 40 min incl. discussion)
  • Time for networking + food + drinks before, in between and after the presentations
  • Talks are held in English
  • We will be taking photos and/or film footage at the event. These will be used to share news about our meetups, and to publicize upcoming events.

The line up:

Eva Höck - Self-supervised image denoising with Noise2Void2

Abstract:
In recent years, neural network based image denoising has pushed the limits of what is possible in enhancement and analysis of biomedical microscopy data. Self-supervised denoising methods are applicable to virtually all noisy datasets, even without dedicated training data being available. Arguably, this facilitated the fast and widespread adoption of machine learning based denoising throughout the life sciences. Noise2Void (N2V) is a self-supervised denoising method which employs blind-spot masking during training in order to learn pixel noise removal. Unfortunately, the blind-spot training can lead to rather visible checkerboard artifacts, thereby reducing the quality of final predictions considerably. In this talk, we present Noise2Void2 (N2V2), our findings how to modify the vanilla N2V setup that help to reduce the unwanted artifacts considerably: an adapted network architecture and novel replacement strategies for the blind spot masking. We validate our modifications on a range of microscopy and natural image data and show that both proposed modifications push the current state-of-the-art for fully self-supervised image denoising.

Bio:
Eva Höck is a Machine Learning Research Scientist in Corporate Research and Technology at Carl Zeiss AG, Munich. Among other fields, she works on Deep Learning for image analysis and quality enhancement in domains such as life sciences and medical image data, with a special focus on 2D and 3D denoising. In her daily work, she uses cutting edge research to generate impact in ZEISS’s diverse product portfolio. She studied at the University of Innsbruck and the Technical University of Munich, where she obtained the MSc Degree in Mathematics in Data Science.

Vladimir Korzinov - Machine Learning Operations. How do you do it?

Abstract:
In this talk I would like to discuss with the audience the hyped topic of MLOps and ML Engineering. The idea is to share our journey to go beyond proof-of-concept status, following a fully functional development, operationalization, and automation cycle. We will talk about team roles, tech stack and organizational challenges.
Bio:
Vladimir is a Lead Machine Learning Engineer in Zeiss Digital Partners at Carl Zeiss AG, where he focuses on building and operating machine learning solutions and supervising the team of data scientists and ML engineers. His journey into AI started with his master’s degree from Delft University, NL and continued in the industry including consulting and retail in energy sector where he was building end-to-end cloud ML systems. In his free time, he geeks out with Raspberry PIs and enjoys hiking and travelling.

Prof. Dr. Martin Spindler - Causal Machine Learning with DoubleML​
Abstract:
While Machine Learning methods have been developed for prediction tasks, many interesting problems in industry and research are causal questions. Examples are pricing, personalized marketing, production optimization, resource allocation and many more. Using Machine Learning for Causal Inference has been an active field in the recent years. In this talk the double ML approach and the open-source library DoubleML are presented. DoubleML is implementing the double machine learning framework of Chernozhukov et al. (2018) for a variety of causal models. In talk also case studies will be presented. Source code, documentation and an extensive user guide can be found at https://github.com/DoubleML/doubleml-for-py and https://docs.doubleml.org.
Bio:
Martin Spindler is Professor of Data Science, Statistics & Econometrics, at the University of Hamburg. In his research he works on the theory and practical applications of Machine Learning and AI, in particluar Causal Machine Leraning. He studied at the University of Regensburg and University of Munich, obtained his PhD from the University of Munich and is a visiting scholar at MIT on a regular basis. He has co-founded Economic AI to support companies to develop state-of-the-art solutions to gain a competitive advantage. The applications comprise amongst others Dynamic Pricing, Financial Forecasting and Planning, advanced A/B Testing and targeted Marketing.

Verwandte Themen

Artificial Intelligence
Machine Learning
Big Data
Predictive Analytics
Open Source

Das könnte dir auch gefallen