Asja Fischer
Asja Fischer researches probabilistic models, deep learning theory and machine learning security. Her 2014 doctorate analyzed the training of restricted Boltzmann machines, including bias in sampling approximations and methods for better learning. With Christian Igel, she wrote an introduction to these models and their training algorithms. After work in Montreal and Bonn, she became a junior professor at Bochum’s security institute in 2018. Her later chair studies the foundations of deep learning, uncertainty, adversarial robustness and risks from generated media.
- Now at
- Ruhr University Bochum
- Works
- 118
- Citations
- 4,774
- h-index
- 27
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2014
Analyze probabilistic network training
Fischer’s doctoral thesis studied restricted Boltzmann machines. It analyzes how sampling approximations can distort learning, proposes faster mixing and studies a parameterization intended to improve learning and robustness to data representation changes.
-
2014
Explain restricted Boltzmann machines
With Christian Igel, she published an introduction to training these stochastic neural networks. The tutorial develops graphical-model and sampling concepts, discusses contrastive divergence and parallel tempering, and uses experiments to show important training behavior.
-
2018
Join Bochum’s security institute
Fischer joined the Horst Görtz Institute in April 2018 as a junior professor for machine learning. She previously held a postdoctoral position in Montreal and an assistant professorship in smart data analysis at Bonn.
-
2019 onward
Deep learning foundations and security
She established her chair in 2019. Her research includes theoretical foundations, uncertainty quantification, adversarial robustness and risks from artificially generated media. These topics connect the learning process with the reliability of deployed models.
Sources and credits
Photo: ELLIS profile; creator not stated. License: Not stated by source. Source: ellis.eu. Cropped to a square and resized. Ask us to remove or replace this photo.