This event is open to the Stanford community
AIMI Postdoc, Joseph Paul Cohen, will provide updates that will cover the 3 papers below that will be presented at ICLR and MIDL this year. They are focused on neural network prediction explanation as well as incorporating metadata to improve model performance.
- Saliency is a Possible Red Herring When Diagnosing Poor Generalization
- Gifsplanation via Latent Shift: A Simple Autoencoder Approach to Progressive Exaggeration on Chest X-rays
- Benefits of Linear Conditioning for Segmentation Using Metadata
Joseph Paul Cohen is a Postdoctoral Fellow at the Center for Artificial Intelligence in Medicine & Imaging at Stanford. Before that Joseph was a Postdoctoral Fellow at Mila and the University of Montreal. Joseph is currently focusing on the limits of AI in medicine with respect to computer vision, genomics, and clinical data. He holds a PhD Degree in Computer Science and Machine Learning from the University of Massachusetts Boston. Joseph has worked on issues related to ML deployment in healthcare focusing on out-of-distribution detection and the limits of generalization. As well as general biology tools for mRNA/DNA representation learning from RNA-Seq and cell counting from microscopy data. Joseph received a U.S. National Science Foundation Graduate Fellowship as well as an IVADO Postdoctoral Fellowship. Joseph is the director of the Institute for Reproducible Research which is dedicated to improving the process of scientific research using technology.