Re: back to the roots

["G. William (Bill) Walster" <bill@xxxxxxxxxxx>]

I wrote "it *can* be a huge mistake".  In my experience, it almost 
always is because few people take the time and spend the effort to 
perform sensitivity analyses (for example using Monte-Carlo simulation) 
to see how sensitive computed answers are to input uncertainties.  Our 
big advantage is supposed to be that we can get rigorous bounds on the 
sensitivity of computations to input uncertainties and do so more 
efficiently.

On 7/1/13 5:24 AM, Vincent Lefevre wrote:
> On 2013-06-29 07:16:51 -0700, G. William (Bill) Walster wrote:
> > I agree that in situations, such as your example, when you are evaluating a
> > monotonic function over an interval and therefore can use endpoint
> > evaluation, extra precision can be useful.  However, surely, this is not the
> > general case.  It can be a huge mistake to use input numbers that are only
> > good to 4 or 5 decimal digits of accuracy and then interpret them as
> > infinitely precise, as Ulrich himself did in his note about the steam
> > turbine that blew up and killed 6 people.
> It isn't necessarily a mistake to regard such inputs as exact, as
> long as at the end, one remembers that they weren't. But wanting to
> get exact results from them (such as with the EDP) is a bad idea in
> general, since it takes resources (time and memory) and is not really
> useful.