Search
Now showing items 1-1 of 1
Smoothness of Functions Learned by Neural Networks
Hladkost funkcí naučených neuronovými sítěmi
bachelor thesis (DEFENDED)
Advisor: Musil, Tomáš
Date Issued: 2020
Date of defense: 07. 07. 2020
Faculty / Institute: Matematicko-fyzikální fakulta / Faculty of Mathematics and Physics
Abstract: Modern neural networks can easily fit their training set perfectly. Surprisingly, they generalize well despite being "overfit" in this way, defying the bias-variance trade-off. A prevalent explanation is that stochastic ...