061: Interpolation, Extrapolation and Linearisation (Prof. Yann LeCun, Dr. Randall Balestriero)
Machine Learning Street Talk (MLST)
Navigating Latent Spaces and Neural Approaches
This chapter investigates how pre-trained neural networks, particularly ResNet classifiers, shape the structure of latent spaces, highlighting the effects of training settings and regularization techniques on information distribution. It also examines the integration of spline functions in deep learning and discusses the innovative concept of Neural Decision Trees, emphasizing their utility in high-dimensional data interpretation. Further, the chapter explores the interplay between continuous and discrete systems in neural networks, raising questions about replicating human cognitive processes in complex tasks like chess.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
