Talk:Grönwall's inequality

I believe there's an error in the "Integral form for continuous functions" section Wikigila (talk) 21:33, 1 August 2010 (UTC)

in the PDE book I saw this in, it didn't restrict $$\!\ \phi(t)$$ to be nonnegative. the proof from PlanetMath doesn't carry over, I don't think.

(The above was not signed by its author.)

You're right, this assumption (as well as the assumption that $$K$$ is positive) is superfluous. These assumptions are in the original and commonly quoted, however, so I'll let it stand for now. Perhaps, if we can quote a good reference, we can drop these assumptions. Or I could provide a proof that works for the more general case, but I am not sure if this is appropriate. Probably not. By the way, the inequality is at least as much Bellman's as Grönwall's. I have edited the page accordingly, with references. And I removed a totally superfluous constant $$L$$ from the statement. Hanche 14:53, 24 April 2007 (UTC)

Err, what the heck, I'll outline a proof here. The differential form, as given in the main article, is immediately proved by differentiating


 * $$\phi(t)\exp\left(-\int_{t_0}^t \psi(s)\, \mathrm{d} s\right)$$

and getting a nonpositive result, thanks to the hypothesis.

The general form follows by applying the differential form to


 * $$\eta(t)=K+\int_{t_0}^t \psi(s)\phi(s) \, \mathrm{d} s $$

which satisifies a differential inequality which follows from the hypothesis (we need $$\psi(t)\ge 0$$ for this; the first form is in fact not correct otherwise). The conclusion from this, together with the hypothesis once more, clinches the proof. No need for nonnegativity of either $$\phi(t)$$ or $$K$$. Hanche 15:15, 24 April 2007 (UTC)

I added the proof outline to the article. If you disagree with this, please discuss here before removing it. But note my rationale for the edit: The proof outline is really very tiny. And it would be hard to decide which if myriads of references (textbooks mostly) to cite. The primary references, which are given in the article, are more of historical interest than as proof references. Although proofs should generally be avoided in articles, I think this is a worthy exception. Divergent viewpoints, anybody? Hanche 14:44, 22 July 2007 (UTC)

Too much detail in the proofs
I think that User:Schmock has added too much detail in his proofs. Wikipedia is not a textbook; it is an encyclopedia, and detailed proofs don't belong, in my opinion. Proof outlines, as they existed before, are a different matter: Particularly in this case, where many textbook proofs are too involved and insufficiently general. In particular, I think that including the standard proof that having a non-positive derivative implies the function being non-increasing is going completely overboard. Well that's my opinion anyway. What do other people think? Hanche (talk) 17:53, 28 March 2008 (UTC)
 * Yep, I agree. I suggest that the verbose proofs be cut out. Oleg Alexandrov (talk) 02:56, 29 March 2008 (UTC)


 * Of course, I like the current version, but I am biased here. Grönwall's inequality is a technical topic, so I think the page is allowed to be technical. The proof for the differential form is so detailed because I wanted to make clear that continuity of the derivative is not needed. While outlines are good, I also like to find complete proofs, because not everyone has the knowledge, training or time to come up with these given the outline. As long as proofs are clearly marked, everyone who wants can easily skip them. When I am in doubt, I prefer to include more details and links than less, because I have no information about the knowledge of the reader. A complete, easy to understand proof also makes it easier to check and verify, that a given statement is correct (important for Wikipedia). Also we can search for and communicate the "best"  (whatever this means in a particular case) proof in this way, which is something of value beyond the correct mathematical statement. References to good textbook proofs (of Grönwall's inequality for this article) are welcome and faster to include in Wikipedia than full proofs. However, not everyone has easy access to a mathematical library. How about adding applications of Grönwall's inequality to this article?  Schmock (talk) 00:24, 31 March 2008 (UTC)
 * I think applications are very important indeed. Perhaps the detailed proofs could be relegated to the end of the article, in a sort-of-appendix, and put the applications before that. Oleg Alexandrov (talk) 02:14, 31 March 2008 (UTC)
 * There's a hiding ("spoiler") function you can use. // Jens Persson (193.10.104.126 (talk) 10:30, 4 September 2008 (UTC))

Integral form with locally finite measures: additional assumption may be needed.
I believe that the additional assumption of continuity on the measure $$ \mu $$ is needed. That seems to be implicit in the proof provided as can be seen in the zero contribution of the integral along diagonals. indydiaz (talk) —Preceding undated comment added 23:32, 5 September 2018 (UTC) What is a "continuous measure"? The current version doesn't make sense! 84.56.144.25 (talk) 15:46, 23 February 2021 (UTC)