Re: AD matrix operations

[Brad Bell]

The type a_double in Octave (and other languages)
    http://www.seanet.com/~bradbell/cppad_swig/a_double_ctor.htm
is a scalar type that will record floating point operations with the 
purpose of creating a corresponding function object
    http://www.seanet.com/~bradbell/cppad_swig/a_fun_ctor.htm

In python, you can create a numpy matrix of type a_double and perform 
operations, like matrix addition, multiplication, etc on the objects. 
The actual matrix loops are done in C (not in python). Can one do  a 
similar thing in Octave ?



On 01/26/2017 07:20 AM, John W. Eaton wrote:
> On 01/26/2017 08:50 AM, Brad Bell wrote:
> > In python, numpy will provide matrix operations for overloaded types.
> > This worked for AD matrix operations; see
> >     http://www.seanet.com/~bradbell/pycppad/pycppad.htm
> > Is there a package, similar to numpy, that works for Octave ?
>
> In Matlab, you can create a function like this (the @double directory 
> needs to be somewhere in your MATLABPATH):
>
>   @double/mtimes.m:
>   function r = mtimes (x, y)
>     fprintf ('my mtimes!\n');
>     r = builtin ('mtimes', x, y);
>   end
>
> to overload the builtin mtimes function that is called for the syntax 
> "x * y" when x and y are double precision numeric objects (that 
> includes sparse and complex).  You'd need a separate definition for 
> single.
>
> Is that what you are looking for?
>
> The overloading feature is not yet implemented in Octave for built-in 
> types like double and single.
>
> jwe