Together with my students and collaborators, I work at the intersection of uncertainty quantification, robustness and interpretability to understand and detect failure modes of deep neural networks.
Our mission is to develop methods, standards and software for AI auditing that will eventually allow the reliable application of AI technology even in high-stakes applications such as medicine.
For that purpose, I co-chair a group of more than 30 contributors from across the world working on data and AI solution assessment methods at the ITU/WHO Focus Group on Artificial Intelligence for Health (FG-AI4H) and co-organize a growing, open research network at aiaudit.org.
If you are interested to collaborate I invite you to take a look here.
|Aug 30, 2021||The site for our Lens to Logit project, a framework to address camera hardware-drift, is up complete with code and data. Learn more|
|Aug 25, 2021||Our paper on uncertainty quantification, interval neural networks and failure mode detection has appeared in International Journal of Computer Assisted Radiology and Surgery. Learn more|
|Aug 20, 2021||I am co-organizing ML4H 2021. Learn more|
|Aug 18, 2021||I am guest-editing a special collection on “Machine Learning for Health: Algorithm Auditing & Quality Control” in the Journal of Medical Systems. Submit your work here|
|Aug 15, 2021||The second iteration of ML4H trial audits has started. Learn more|
- IJCARSDetecting failure modes in image reconstructions with interval neural network uncertaintyInternational Journal of Computer Assisted Radiology and Surgery 2021
- ICLRPost-Hoc Domain Adaptation via Guided Data HomogenizationIn ICLR 2021 Workshop on Robust and Reliable Machine Learning in the Real World Workshop (RobustML) 2021
- ITU/WHOGood practices for health applications of machine learning: Considerations for manufacturers and regulatorsIn Proceedings of the ITU/WHO Focus Group on Artificial Intelligence for Health (FG-AI4H) - Meeting K 2020
Top 10%ML4H Auditing: From Paper to PracticeIn Proceedings of the Machine Learning for Health NeurIPS Workshop 2020
Top 10%Detecting Failure Modes in Image Reconstructions with Interval Neural Network UncertaintyIn ICML 2020 Workshop on Uncertainty & Robustness in Deep Learning 2020