About this document

Can Stable and Accurate Neural Networks Be Computed? - On The Barriers of Deep Learning and Smale by liyuqi14ab14 is a document available to read on EtoBox.

The document explores the paradox of instability in deep learning (DL) methods despite the theoretical existence of stable neural networks (NNs) with good approximation qualities. It presents results indicating that while stable NNs can exist for well-conditioned problems, no algorithm can reliably compute them under certain conditions, highlighting barriers in achieving both stability and accuracy. The authors introduce Fast Iterative REstarted NETworks (FIRENETs) as a solution that demonstrates stability

Author

liyuqi14ab14

Language

EN