DSLT 2. Why Neural Networks obey Occam's Razor

Liam Carroll

2023年6月18日

Published on June 18, 2023 12:23 AM GMTTLDR; I synthesise why Watanabe's free energy formula explains why neural networks have the capacity to generalise well, since different regions of the loss landscape have different accuracy-complexity tradeoffs. I also …

TLDR; I synthesise why Watanabe's free energy formula explains why neural networks have the capacity to generalise well, since different regions of the loss landscape have different accuracy-complexi… https://www.lesswrong.com/posts/CZHwwDd7t9aYra5HN/dslt-2-why-neural-networks-obey-occam-s-razor