Re: directional derivative,--



David C. Ullrich wrote:
On Tue, 24 Apr 2007 01:43:40 +0100, JEMebius <jemebius@xxxxxxxxx>
wrote:

What about the well-known rectangular circular half-cone,
the graph of (x, y) -> z = sqrt(x^2 + y^2)?

That has no directional derivative in _any_ (non-zero) direction (at the origin, which is presumably the point you're talking about.)

At least not according to what I've always thought was the
standard definition, as at

http://en.wikipedia.org/wiki/Directional_derivative



That is absolutely correct, if it is indeed standard to consider only entire lines through the point in question. Is it standard in university and college math curricula? - I don't know.

When defining mathematical concepts I want to stay as closely as possible to everyday physical reality. So I identified "direction" with "half-line" rather than with "line".

In my opinion it is partly a matter of convention, partly a matter of to what degree one wants to change the common meaning of an everyday word into a precise mathematical meaning, and partly a matter of convenience whether one considers half-lines or entire lines through the point in question.

Johan E. Mebius


At first sight the problem seems underspecified? incompletely posed? badly posed? provocative, i.e. aimed at putting in motion the reader's creativity?

Or did the author mean to ask:
"Prove that no real-valued function f ==that has a continuous derivative at a point C==
has a positive directional derivative at C for every possible direction u."

BTW, the paraboloid (x, y) -> z = x^2 + y^2 does not qualify because the slope at (0, 0) is zero in every direction.
This example shows that zero local slope is not in contradiction to positive average slope.

Ciao: Johan E. Mebius


chrizm7@xxxxxxxxx wrote:
A problem asks me to show that no real-valued function f has a
positive directional derivative at a fixed point c for every possible
direction u.

But wouldn't a paraboloid work? Fix c corresponding to the vertex.
Then for any direction u, the slope will always be positive
(increasing).

To be clear, the exact problem is worded: Prove that there is no real-
valued function f such that f'(c;u) > 0 for a fixed point c in R^n and
every nonzero vector u in R^n.



************************

David C. Ullrich
.

Relevant Pages

  • Re: directional derivative,--
    ... David C. Ullrich wrote: ... if it is indeed standard to consider only entire lines through ... This example shows that zero local slope is not in contradiction to positive average slope. ...
    (sci.math)
  • Re: The definite integral.
    ... >David C. Ullrich wrote: ... >> You've said a few times that one can prove ftc without mvt. ... Note that the slope of this line is not 0. ...
    (sci.math)
  • Re: FO logic without equality
    ... David C. Ullrich wrote: ... S u EQ_AX satisfiable iff S normal satisifiable ... FOL with or without equality does not matter to ...
    (sci.logic)
  • Re: What is the 1st order formal system known as PA?
    ... >> many different interpretations. ... > how can we be so sure it won't matter to _all_ formulae of L? ... >> David C. Ullrich ...
    (sci.logic)
  • Re: What is the Result from Invoking this Halt Function?
    ... >The halt analyzer is not required to answer every question that is ever ... >asked no matter who is doing the asking. ... >> David C. Ullrich ...
    (comp.theory)