perfect!

thanks

but in general I wont know where to start and so the local search wont yield a better solution.

so the min of global search and local search will usually be the min found by the global search which in this case is wrong.

Hi Pagan

Thanks for the suggestion. Wont this only work though if I know the solution (i.e. a start the local search near the correct answer). In this case -yes I know the answer, but in general I wont.

It seems very odd the maple choses to ignore a better solution that it "knows" about.

thanks 

never heard of system/ ssystem -I'll look into it now. thanks very much for your help.

sorry I've just realised that my question wasnt very specific...

I would like to "reset" the memory. One potential way forward seems to be invoking garbage collection every few loops. According to the help pages the garbage collection deletes "data to which no reference is made". What exactly does "no reference" mean? There wont be a variable that isnt mentioned (referenced?) in the code will there? Or does "referenced" mean something different?

Is garbage collection the solution to my problem and if so exactly how does it work?

THANKS!

Hi Joe

the first version works very well but I couldn't get the numerical version to work (get error message -I'm using Maple 11). What is the option `parameters' that you use in dsolve?

thanks

Hi acer

if you could have a look at this when you have a minute that would be great. I appreciate that not one piece of software is best for everything but for some reason the toolbox is very poor  -I suspect I must be doing something wrong. The fitted solution really bears no resemblance to the data!

I will change the method to multistart and let you know how it goes.

I've got the infolevel set to 10 so I can see what is going on .

Scop you do compile before running.

thanks for your help!

Hi Joe

thanks for the reply but isnt "Statistics:-NonlinearFit" a local fit? I'd be worried about local minima for more complicated problems.

Thanks David

Is there a way to specify shrinking of the array as a command or do you just mean drag the black double headed arrows? Dragging the arrow seems to do one row or one column at a time rather than the whole array so it will be a bit hit or miss.

yeah with new code I now use LinearAlgebra rather than linalg. But some of these procedures were written a while ago using linalg so I just did the whole thing with linalg rather than transfer everything over to LinearAlgebra- in my experience its not a particularly smooth transition!

But although LinearAlgebra would be ideal I'd be surprised if moving to it would resolve the problem.

Is there no way to free up memory as this is running. Each loop stands alone -it doesnt need results from previous loops so I don't understand why memory usage is increasing with loops.

thanks!

I am having exactly the same problem with lost kernel connection during a long run.

and I have also just upgraded my memory, though until now I hadnt linked the two...

if anyone has an opinion about what is going on I would love to know

thanks!

Hi Stephanie

I've already tried Adept Scientific. They don't run courses themselves and werent able to suggest anyone who does. Can you think of any other possibilities?

thanks

Hi Acer

yes im using procedures, By outputting I mean writing the matrix to a file (using fprint) and none of those results are needed later.

Would be great if I could email you the code (though I warn you it is a mess -it has evolved to do many more tasks than originally planned so its a bit of a mess). do you have an email address that I could use?

thanks!

I'm afraid the code is long and rambling and theres no one bit that is particularly representative. I don't know enough to know what is most likely to be contributing to the excessive memory usage.

Is a large matrix likely to be a problem -I'm currently storing results in a single (v large matrix) and outputting periodically. If I didn't store results and just outputted reults each loop is that likely to be a significant saving. I realise this is difficult to answer without seeing the code. But in general what eats up memory?

thanks

Hi

Is there anyway to get around this memory problem? Is there an option to allocate more to maple. or alternatively after 1000 runs or so can I automate saving the important variables that I need and then "restart" (whilst running the code) to clear the memory?

thanks