Skip to main content
15 events
when toggle format what by license comment
S Dec 18, 2020 at 17:02 history bounty ended CommunityBot
S Dec 18, 2020 at 17:02 history notice removed CommunityBot
Dec 14, 2020 at 19:39 vote accept ACR
Dec 11, 2020 at 20:08 history edited ACR CC BY-SA 4.0
added 545 characters in body
Dec 11, 2020 at 12:15 comment added Dieter Kadelka I'm wondering why the focus in this question is on inherently numerical instable algorithms for numerical differentiation. What is bad with f.i. using cubic splines? I don't understand.
Dec 10, 2020 at 22:14 comment added Steven Gubkin @AllanMacLeod Indeed, it would almost certainly be better to fit some curve to the data points, and then talk about the derivatives of the model.
Dec 10, 2020 at 22:06 answer added Federico Poloni timeline score: 2
Dec 10, 2020 at 20:13 answer added Carlo Beenakker timeline score: 2
Dec 10, 2020 at 18:17 comment added Dirk No one talks about it because there is no definitive way here. One can, however, make smoothness assumptions and use Taylor expansion to prove error bounds. This should be in most books on numerical math. Without assumptions on f nothing meaningful can be said.
S Dec 10, 2020 at 15:42 history bounty started ACR
S Dec 10, 2020 at 15:42 history notice added ACR Authoritative reference needed
Dec 8, 2020 at 20:28 comment added ACR Okay, good to know that numerical differentiation is considered to be an old problem among mathematicians. But despite its shortcomings, it is still widely used. If we use the forward difference formula then the first derivative at x is which x? Nobody talks about it clearly.
Dec 8, 2020 at 18:26 comment added Allan MacLeod Should point out that numerical differentiation is a classic ill-posed problem, meaning that small errors can cause quite large effects.
Dec 8, 2020 at 17:13 history edited ACR CC BY-SA 4.0
added 5 characters in body; edited title
Dec 8, 2020 at 15:33 history asked ACR CC BY-SA 4.0