Explore transfer learning in deep learning with a focus on medical imaging applications in this 46-minute lecture by Maithra Raghu from Cornell University and Google Brain. Delve into the fundamentals of transfer learning, its applications in medical imaging, and the evaluation of transfer performance. Examine the relationship between ImageNet model performance and transfer capabilities, analyze chest X-ray results, and learn about Canonical Correlation Analysis (CCA) for feature similarity assessment. Investigate the similarity of deep representations and feature reuse in transfer learning. Conclude by considering open questions in the field, gaining valuable insights into this crucial area of deep learning research.
Towards Understanding Transfer Learning with Applications to Medical Imaging
Overview
Syllabus
Intro
Transfer in Deep Learning Applications
Github for Transfer Learning
Do better ImageNet models transfer better? (Komblith, Shlens, Le), 2019
Performance Results on Chest X-rays
Evaluating Transfer: Main Takeaways
Going Beyond Performance Evaluations
CCA for Feature Similarity
Similarity of Deep Representations
Feature Similarity in Transfer with CCA Compare feature similarity of
Feature Similarity and Reuse
Open Questions
Taught by
Simons Institute
