Mathematica - Preventing NIntegrate from reevaluating after DumpSave and Get

Author

Preventing NIntegrate from reevaluating after DumpSave and Get

Chris Rodgers

2006-08-18, 4:07 am

Hi,

I have been running a computation in Mathematica which attempts to
NIntegrate[myformula,{t,0,infinity}] for many different model parameters.

The results are in a large list. e.g.
resultslist = {0, 7, 13.33,
NIntegrate[myformula_at_tricky_parameter
_vals,{t,0,infinity}],11,2,...}

Notice that sometimes NIntegrate does not converge for particular model
parameters.

At the end of a long-running batch job, I DumpSave this results list.

Now, my question is this:

When I load the results back into Mathematica and try to use them things
are very slow. I suspect that this is because Mathematica retries
evaluating all the NIntegrate[...] on which it got stuck the first time
around.

I have tried e.g.

Get["MyFile.mx"]

sanelist = resultslist /. NIntegrate[___]->NotANumber

but this also takes ages evaluating the NIntegrate[...].

Is there an easy way to load the .mx file and eliminate any of the
non-converged NIntegrate[...] entries without having to wait for them to
all be evaluated another time?

Many thanks in advance,

Chris Rodgers.

albert

2006-08-19, 4:03 am

Hi,

> When I load the results back into Mathematica and try to use them things
> are very slow. I suspect that this is because Mathematica retries
> evaluating all the NIntegrate[...] on which it got stuck the first time
> around.
>
> I have tried e.g.
>
> Get["MyFile.mx"]

even here the NSolves will be evaluated since Mathematica basically tries to
evaluate everything it sees...

> sanelist = resultslist /. NIntegrate[___]->NotANumber

> but this also takes ages evaluating the NIntegrate[...].

to import the Results unevaluated, you could try something like this:

Block[{NIntegrate},
Get["MyFile.mx"]/.NIntegrate[___]->NotANumber
]

this is untested, but I think it should work. Anyway it should be worth a
try.

hth,

albert

Jens-Peer Kuska

2006-08-19, 4:03 am

Hi,

you must remove the NIntegrate[] calls *before* you make the dump-save.
Or, just an idea

Block[{NIntegrate},
NIntegrate[any__]:=Indeterminate
Get["MyFile.mx"]
]

overwrite temporal NIntegrate[]

Regards
Jens

Chris Rodgers wrote:
> Hi,
>
> I have been running a computation in Mathematica which attempts to
> NIntegrate[myformula,{t,0,infinity}] for many different model parameters.
>
> The results are in a large list. e.g.
> resultslist = {0, 7, 13.33,
> NIntegrate[myformula_at_tricky_parameter
_vals,{t,0,infinity}],11,2,...}
>
> Notice that sometimes NIntegrate does not converge for particular model
> parameters.
>
> At the end of a long-running batch job, I DumpSave this results list.
>
> Now, my question is this:
>
> When I load the results back into Mathematica and try to use them things
> are very slow. I suspect that this is because Mathematica retries
> evaluating all the NIntegrate[...] on which it got stuck the first time
> around.
>
> I have tried e.g.
>
> Get["MyFile.mx"]
>
> sanelist = resultslist /. NIntegrate[___]->NotANumber
>
> but this also takes ages evaluating the NIntegrate[...].
>
> Is there an easy way to load the .mx file and eliminate any of the
> non-converged NIntegrate[...] entries without having to wait for them to
> all be evaluated another time?
>
> Many thanks in advance,
>
> Chris Rodgers.
>

Peter Pein

2006-08-19, 4:03 am

Chris Rodgers schrieb:
> Hi,
.....
> I have tried e.g.
>
> Get["MyFile.mx"]
>
> sanelist = resultslist /. NIntegrate[___]->NotANumber
>
> but this also takes ages evaluating the NIntegrate[...].
>
> Is there an easy way to load the .mx file and eliminate any of the
> non-converged NIntegrate[...] entries without having to wait for them to
> all be evaluated another time?
>
> Many thanks in advance,
>
> Chris Rodgers.
>

Hi Chris,

maybe it helps to do a

resultlist = resultlist /. _NIntegrate -> NotANumber

_before_ dumpsaving?

I have to admit, I didn't try it.

Peter

Sponsored Links