061: Interpolation, Extrapolation and Linearisation (Prof. Yann LeCun, Dr. Randall Balestriero)
Machine Learning Street Talk (MLST)
The Importance of Regularization in Neural Network Training
In order for a network to accurately learn, it needs sufficient samples to understand the relevant information. However, even seemingly irrelevant information can be useful in reducing training loss. Therefore, claiming that adding extra dimensions is always beneficial is not necessarily true. It depends on having the perfect regularizer, model, and training regime. Prior knowledge about useless information can be helpful in improving generalization.
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.
