Timeline for What is required for neural network to approximate discontinuous function?

12 events

when toggle format	what		by	license	comment
Dec 31, 2023 at 1:05	comment	added	Ben Reiniger		The paper @John links uses activation functions that depend on the target function, which doesn't seem particularly in the spirit of neutral networks.
Dec 30, 2023 at 20:06	history	edited	Sycorax♦	CC BY-SA 4.0	added 726 characters in body
Dec 30, 2023 at 19:21	history	edited	Sycorax♦	CC BY-SA 4.0	deleted 23 characters in body
Dec 30, 2023 at 19:06	history	edited	Sycorax♦	CC BY-SA 4.0	added 331 characters in body
Dec 30, 2023 at 18:55	history	edited	Sycorax♦	CC BY-SA 4.0	added 131 characters in body
Dec 30, 2023 at 18:09	history	edited	Sycorax♦	CC BY-SA 4.0	Add link to thread
Dec 30, 2023 at 18:03	history	edited	Sycorax♦	CC BY-SA 4.0	Add link to thread
S Dec 30, 2023 at 17:46	history	edited	Sycorax♦	CC BY-SA 4.0	Add link to thread
S Dec 30, 2023 at 17:46	history	suggested	Ggjj11	CC BY-SA 4.0	Add link to thread
Dec 30, 2023 at 17:45	review	Suggested edits
S Dec 30, 2023 at 17:46
Dec 4, 2021 at 14:27	comment	added	John		There is now a proof that a three-layer neural network can approximate any discontinuous function: arxiv.org/abs/2012.03016 However, this does explicitly not say, that there is a learning algorithm that converges to the solution.
Sep 1, 2018 at 1:16	history	answered	Sycorax♦	CC BY-SA 4.0