Interpretability and Out-of-Distribution Generalization in Deep Predictive Models
Explore interpretability and out-of-distribution generalization in deep predictive models through this comprehensive conference talk by Rajesh Ranganath at the Computational Genomics Summer Institute (CGSI) 2022. Delve into cutting-edge research on addressing nuisance-induced spurious correlations in out-of-distribution generalization, examining how interpretability methods can learn to encode predictions in their interpretations, and understanding real-time Shapley value estimation. Gain insights from related papers, including work on FASTSHAP and the analysis of learned explanations in interpretability methods. Enhance your understanding of advanced concepts in deep learning and their applications in computational genomics during this 41-minute presentation.