Für eine korrekte Darstellung dieser Seite benötigen Sie einen XHTML-standardkonformen Browser, der die Darstellung von CSS-Dateien zulässt.

News

Domain Adaptation and Transfer for deep learning models.


Content

While deep learning models have achieved outstanding performances on various tasks, those tasks are mostly performed within a certain domain and deep learning models still lack the ability for out of domain generalization.
Here Domain Adaptation and Transfer comes into play. Domain Adaptation and Transfer aim to train models which are not only able to perform in their trained domain but are also able to solve tasks outside of their trained domain.
The techniques for Domain Adaptation and Transfer vary a lot and have impact on several deep learning topics, such as model robustness or neural architecture search.

Qualification Goals

  • Read, understand, and explore scientific literature
  • Summarize a current research topic in a concise report (10 single-column pages + references)
  • Give two presentations about your topic (3 minutes flash presentation, 15 minutes final presentation)
  • Moderate a scientific discussion about the topic of one of your fellow students
  • Review a (draft of a) report of a fellow student

Schedule

Kickoff-Meeting: October 22nd
Flash-Presentations: December 3rd (tentative)
Final-Presentation: January 31st/ February 7th (tentative)

Literatur

  1. A Fourier-based Framwork for Domain Generalization https://arxiv.org/abs/2105.11120 
  2. FSDR: Frequency Space Domain Randomization for Domain Generalization https://arxiv.org/abs/2103.02370 
  3. Rethinking and Improving the Robustness of Image Style Transfer https://arxiv.org/abs/2104.05623 
  4. Reducing Domain Gap by Reducing Style Bias https://arxiv.org/abs/1910.11645 
  5. What Can Style Transfer and Paintings Do For Model Robustness? https://arxiv.org/abs/2011.14477 
  6. Neural Discrete Representation Learning https://arxiv.org/abs/1711.00937
  7. Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets https://arxiv.org/abs/2107.00860 
  8. Neural Architecture Transfer https://arxiv.org/abs/2005.05859
  9. XferNAS: Transfer Neural Architecture Search [1907.08307] XferNAS: Transfer Neural Architecture Search - arXiv 
  10. Transfer NAS: Knowledge Transfer between Search Spaces with Transformer Agents https://arxiv.org/pdf/1906.08102.pdf