Computer Vision Research Assistant — Duquesne University

Two threads of research came out of this position:

• An interpretability analysis of CNNs applied to genomic data, giving population geneticists a lightweight framework for detecting signatures of natural selection. See the project Selection Inference with CNNs and the paper On CNNs for Selection Inference (PLOS Comp. Bio.).
• A deep-learning framework using learned geometric and higher-order features for image denoising, beating model-based methods on standard metrics. See Learned Regularizers for Image Denoising and the paper Learned Regularizers and Geometry for Image Denoising (BMVC).