Strategic COMET-Project stratDL (2019-2022)

Strategic Research on Fundamentals of Deep Learning Applications

Motivation

The aim of this strategic project is to provide fundamental methodological building blocks necessary to achieve our objectives. We plan research to support the multi-firm projects with foundations to:

  • advance, establish and apply knowledge-based and probabilistic methods in order to increase trustworthiness of machine learning based models in data and image analysis
  • facilitate and support the efficient design, training and adaptation of (deep) machine learning models for data and image analysis by knowledge and model integration
  • devise with best-practice workflows and software innovations to meet specific system requirements and constraints by exploiting interdisciplinary research synergies between data-driven modeling and software engineering approaches

The work conducted in this project will be guided strongly by our scientific partners.

Expected Results

  • Mathematical methods for assessing trustworthiness of deep learning models
  • Mathematical foundations for transfer learning
  • Methods to assess the level of privacy achievable within distributed deep learning systems
  • Machine learning based approaches for causal discovery suitable for industrial applications, especially for the detection of control loops
  • Runtime and convergence guarantees for robust learning algorithms tailored for applications in fault detection and prediction
  • Convergence and performance guarantees: We further aim to increase online adaptation of deep networks, where we expect to provide performance guarantees of online implementations
  • Application specific benchmark datasets for the evaluation of predictive behavioral analytics from the domains of safety control, security surveillance, scouting and tracking based ergonometric analytics
  • Python software prototype for predictive behavioral analytics and its benchmarking 

Scientific Cooperation

  • PhD cooperation with JKU-FLLL (Johannes Kepler University Linz / Department of Knowledge-Based Mathematical Systems) on the topic of Transfer Learning (Prof. Susanne Saminger-Platz)
  • PhD cooperation with JKU-FAW (Johannes Kepler University Linz / Institute for Application Oriented Knowledge Processing) on the topic of Automated Data Quality Engineering (Prof. Wolfram Wöß)
  • Master students cooperation with University of Passau, Chair of Data Science (Prof. Michael Granitzer)
    Master students cooperation with KTH Royal Institute of Technology, School of Electrical Engineering and Computer Science (Prof. Mihhail Matskin)
    Master students cooperation with Maastricht University, Department of Data Science & Knowledge Engineering (Prof. Gerhard Weiß)
  • Master students cooperation with Radboud University Nijmegen, Institute for Computing and Information Sciences (Prof. Tom Heskes)

Funding Partner

This project is subsidized in the frame of COMET – Competence Centers for Excellent Technologies by BMVIT, BMDW, State of Upper Austria and its scientific partners. The COMET program is handled by FFG.

Contact

Bernhard Freudenthaler

Freudenthaler Bernhard

Area Manager Data Science
Phone: +43 50 343 850

back