Converting AD<double> to double about cppad

[bradbell]

As I mentioned, but perhaps was not clear about, all the derivatives of the Value(Var2Par(y)) are zero, even it the derivatives of y are not zero. It you need to track the dependency (derivatives) you need to use an AD type. So the real question is why are you using double and not AD<double> ?

[bradbell]

If the source for the transformation is not to large, convert it to a function where double is replaced by a template parameter so that you can use the source code with both double and AD<double>.

Also, it sounds like (but I am not sure) you are interpolating values, so you may want to look at
https://coin-or.github.io/CppAD/doc/interp_onetape.cpp.htm
https://coin-or.github.io/CppAD/doc/interp_retape.cpp.htm

[vimalrajayyappan]

I found the way. There is a function in CppAd::Value(AD<>) will give the value.

Is that right?

[vimalrajayyappan]

While I was getting my code Compiled. Its showing error while Runtime. Kindly help me with this. The error is

cppad-20160000.1 error from a known source:
Value: argument is a variable (not a parameter)
Error detected by false result for
Parameter(x)

[bradbell]

You can only extract the value for a parameter (a parameter does not depend on the independent variables). This is because the value of an AD object has type Base which does not track dependency on AD independent variables. If you really want the value with out the dependency, use Var2Par;
https://coin-or.github.io/CppAD/doc/var2par.htm

[vimalrajayyappan]

Dear BradBel. Exactly solved the issue..nice

But I got stuck with one more issue. When I'm converting the variable using Var2Par in cost functions, the optimiser is not recognising the cost.. Its not even considering the cost even though the values are high and appropriate . Can you help?

I know this explanation makes no meaning. But I figured it out in that way.

[bradbell]

Why are you using the Value function ?

[vimalrajayyappan]

Actually there are certain implementations where I'm using "double" datatype. Which is causing a typecasting issue. Hence I'm using CppAD::Value to get it to its base type and feed them.

[vimalrajayyappan]

One good example I can give you.

The points getting optimised are in global frame. I'm having a common.hpp file where there is a transformation function using double. I need to use that function to transform points in local frame. Hence typecasting..

But from your answer I get to know we must use Ad. to track. Is tat ryt?

[vimalrajayyappan]

Dear BradBel. Thanks a lot for your kind help. I will dig into the links u have shared. 😊