automatic differentiation using custom ndarray?

Matthew Koichi Grimes mkg at cs.nyu.edu
Thu Nov 9 19:49:02 CST 2006


I'd love to be able to do automatic differentiation using numpy
ndarrays. There are some AD libraries in C++ that work through operator
overloading on custom numeric types (e.g. "adouble"). I have no
experience in in creating custom ndarrays, but would it be a huge
project (or even possible) to make a custom ndarray that used such
augmented numeric types? From reading Jonathan Wang's thread
(http://www.mail-archive.com/numpy-discussion@lists.sourceforge.net/msg03487.html), 

it seemed like ndarrays with custom numeric types are still a work in
progress.

I'm thinking specifically of the ADOL-C library, but if people with
experience with various AD libraries have a better recommendation, I'd
be all ears. I've also looked at CPPAD, but it doesn't seem to operate
nearly as transparently.

-- Matt


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642




More information about the Numpy-discussion mailing list