[Numpy-discussion] memoization with ndarray arguments

Gael Varoquaux gael.varoquaux@normalesup....
Sun Mar 22 10:03:57 CDT 2009


On Sat, Mar 21, 2009 at 07:18:56PM +0100, Paul Northug wrote:
> I would like to 'memoize' the objective, derivative and hessian
> functions, each taking a 1d double ndarray argument X, that are passed
> as arguments to
> scipy.optimize.fmin_ncg.

I have develop a library that does this. It uses heavy, ugly tricks. It
is called joblib, 
http://pypi.python.org/pypi/joblib
You can use it, I use it all the time. It is documented and tested (I
just uploaded a new version with better documentation).

Now I believe that this implementation is not the right approach to solve
my problem, and I hope to find some time to do fairy radical changes to
joblib. This is one of the reason why I haven't been advertizing it much.
However, it does probably what you want, or at least partly.

Cheers,

Gaël


More information about the Numpy-discussion mailing list