Abstract: 

Extracting insights from imaging data used to be straightforward: every component of imaging systems was engineered by humans, the analysis and interpretation of the collected data was driven by human understanding and experience, and only humans were responsible for the impact of the decisions stemming from such insights. Today, however, machine learning permeates every stage of image acquisition and analysis, challenging our understanding of robustness, interpretability, and fairness in automated decision making. This talk will present three approaches to leverage the power of data-driven models for imaging applications while increasing their trustworthiness, focusing on biomedical imaging. These approaches enable precise mathematical claims about what modern networks compute in the context of inverse problems (via learned proximal networks), facilitate efficient and rigorous testing for interpretable concepts for classification problems (via testing by betting), and ensure compliance with fairness guarantees even in incomplete-data regimes (via proxy attributes).

Zoom Link (if unable to attend in-person): https://upenn.zoom.us/j/93115449335