Ok, thanks Robert. An optional keyword parameter to Statistics:-NonlinearFit might do, such as {globalsearch::truefalse:=false}. The interactive Optimization assistant has something like this, a checkbox for global searching (applied to NLP solving, say). So it would be nice if the non-interactive components which use Optmization, such as Fit and NonlinearFit, could also get such options.

acer

@Christopher2222 You are not just imagining it. Some navigation in the new site is slower than it was.

@slowlai That specfunc(anything,min) part of that frontend invocation is there so that min() calls don't get frozen entirely (which would prevent convert/piecewise from being to see & operate on its contents). As used, that frontend invocation will freeze most else, including products such as x*lambda. The idea is to treat x*lambda as if it were a single name, by freezing it with frontend, so that convert/piecewise will function on it.

Another way could be to freeze just that subexpression x*lambda, "by hand",

expr:=min(x*lambda,0);
ice:=freeze(x*lambda);
convert(subs(x*lambda=ice,expr),piecewise,ice);
thaw(%);

@slowlai That specfunc(anything,min) part of that frontend invocation is there so that min() calls don't get frozen entirely (which would prevent convert/piecewise from being to see & operate on its contents). As used, that frontend invocation will freeze most else, including products such as x*lambda. The idea is to treat x*lambda as if it were a single name, by freezing it with frontend, so that convert/piecewise will function on it.

Another way could be to freeze just that subexpression x*lambda, "by hand",

expr:=min(x*lambda,0);
ice:=freeze(x*lambda);
convert(subs(x*lambda=ice,expr),piecewise,ice);
thaw(%);

@Will "We think this is the easiest and fastest way to get rich math onto the web; simply compose your post in Maple. Save the worksheet and upload it to Primes."

Except that is not easy at all. It's far too much effort, to fire up Standard, compose, save to a new file with a new name, then open (and find) and load that up into Primes, all just in order to submit a reply to a post.

The Editor's menubar's "Maple Math" button is thus the best (only!?) hope for a fast method, then, given what you say about the clipboard.

But the displayed math graphic due to using the "Maple Math" button ought then to be aesthetically acceptable, and so far members are all agreeing that what comes from using the "Maple Math" button is ugly. See Erik's post above, which shows such.

And I cannot stress how important 1D Math alt tags are for anything representing 2DMath input. Examples of code in posts which cannot be copied from Primes and inserted elsewhere (as 1D input at least) are very close to having no utility. I hope that the .mw translation you mentioned provides that throughout.

Let me try an example, using firefox 3.0.7 on Windows XP. The expression that I type into the Maple Math box is,

   BesselJ(0,x) - Theta(psi) - GAMMA(4*Xi/3)

Note the (valid and legal Maple syntax for both 1D and 2D input) spaces between subterms and minus signs. If I check the Evaluate expression box then the popup has an error message saying I have entered an invalid expression, which is a bug. So I uncheck that box. See below, to what gets produced. Of course, apart from the +-+ bug there is also the aesthetic problem that people have mentioned.

Now I'll try it without the (valid!) spaces. It's just not pretty. It does have 1D code alt tags, which is good.

That last one is an image. So the mapleprimes server could make a prettier image, yes? (Why not use a browser/font rendering solution, btw? Images seem anachronistic. How about STIX?)

@Will "We think this is the easiest and fastest way to get rich math onto the web; simply compose your post in Maple. Save the worksheet and upload it to Primes."

Except that is not easy at all. It's far too much effort, to fire up Standard, compose, save to a new file with a new name, then open (and find) and load that up into Primes, all just in order to submit a reply to a post.

The Editor's menubar's "Maple Math" button is thus the best (only!?) hope for a fast method, then, given what you say about the clipboard.

But the displayed math graphic due to using the "Maple Math" button ought then to be aesthetically acceptable, and so far members are all agreeing that what comes from using the "Maple Math" button is ugly. See Erik's post above, which shows such.

And I cannot stress how important 1D Math alt tags are for anything representing 2DMath input. Examples of code in posts which cannot be copied from Primes and inserted elsewhere (as 1D input at least) are very close to having no utility. I hope that the .mw translation you mentioned provides that throughout.

Let me try an example, using firefox 3.0.7 on Windows XP. The expression that I type into the Maple Math box is,

   BesselJ(0,x) - Theta(psi) - GAMMA(4*Xi/3)

Note the (valid and legal Maple syntax for both 1D and 2D input) spaces between subterms and minus signs. If I check the Evaluate expression box then the popup has an error message saying I have entered an invalid expression, which is a bug. So I uncheck that box. See below, to what gets produced. Of course, apart from the +-+ bug there is also the aesthetic problem that people have mentioned.

Now I'll try it without the (valid!) spaces. It's just not pretty. It does have 1D code alt tags, which is good.

That last one is an image. So the mapleprimes server could make a prettier image, yes? (Why not use a browser/font rendering solution, btw? Images seem anachronistic. How about STIX?)

@epostma "..not very useful". That's true, since it is not useful.

@epostma "..not very useful". That's true, since it is not useful.

Thanks Alec. I never noticed this old response, until I started reviewing posts to add tags. Sigh.

At the time, I figured just as you have done, that the problem is with Matrix type checks being wrongly evaluated (like Matrix constructor calls). But I don't see a reason why it can't be fixed inside CodeGeneration, as I'm not sure that fixing it would interfere with any part that's working properly at present.

The reason I want this,by the way, is that if working I'm pretty sure that I would be able to cobble together an extension to CodeGeneration[C] which "undertstood" large parts of the LinearAlgebra package for float[8] rtables, and which could emit C code with appropriate calls to the BLAS or CLAPACK equivalents. But I can't really begin properly, in light of this bug.

acer

Thanks Alec. I never noticed this old response, until I started reviewing posts to add tags. Sigh.

At the time, I figured just as you have done, that the problem is with Matrix type checks being wrongly evaluated (like Matrix constructor calls). But I don't see a reason why it can't be fixed inside CodeGeneration, as I'm not sure that fixing it would interfere with any part that's working properly at present.

The reason I want this,by the way, is that if working I'm pretty sure that I would be able to cobble together an extension to CodeGeneration[C] which "undertstood" large parts of the LinearAlgebra package for float[8] rtables, and which could emit C code with appropriate calls to the BLAS or CLAPACK equivalents. But I can't really begin properly, in light of this bug.

acer

@Alec Mihailovs Thanks, Alec!

> m - (m-n+1) + 1;
                                       n

I guess I had it producing 2 more values than it had to. Still "correct"... but wasteful.

I almost missed seeing this response, sorry. I find it hard to not miss some responses, as the Recent page doesn't have results sorted by Last Action times.

@Alec Mihailovs Thanks, Alec!

> m - (m-n+1) + 1;
                                       n

I guess I had it producing 2 more values than it had to. Still "correct"... but wasteful.

I almost missed seeing this response, sorry. I find it hard to not miss some responses, as the Recent page doesn't have results sorted by Last Action times.

Hi Joe,

datatype=anything would be acceptable, if it is unsure about whether overflow might occur, although I think that I might prefer integer[4] and damn the torpedoes. Returning a float result looks plain wrong to me.

acer

Hi Joe,

datatype=anything would be acceptable, if it is unsure about whether overflow might occur, although I think that I might prefer integer[4] and damn the torpedoes. Returning a float result looks plain wrong to me.

acer

Two degrees of freedom are at work: the cumulative score, and the number of participants voting.

The nuclearphynance site shows both of those explicitly, which is fine and good.

The current mapleprimes scoring mechanism shows both as well, although both are implicit. But that's OK. Most of us can easily add the two numbers, up--votes and down-votes, to get the number of participants, and subtract them to get the cumulative score.

Any scheme in which a cumulative single score is shown, while the number of participant is not shown or deducible, is suboptimal.

The Application Center's ranking scheme falls into that last category.

I can't tell whether or not Jacques is suggesting merging the up- and down-votes and keeping or retaining implicit display of the number of participants. I'd be surprised if he were suggesting ditching any and all implicit display of the total number of voters.

acer