[Ipopt] Question about Jacobian approximation: option finite-difference-values

[Atri Dutta]

Hi, 

I am using IPOPT with Matlab interface. When I provide exact derivatives for Jacobian and Hessian functions, the program runs fine and solves the problem I am working on. Using exact derivatives for the Jacobian and approximation for the Hessian also work out fine, but I am having trouble when I try to approximate the Jacobian. 

IPOPT throws an error "You must provide a function that returns the first derivatives (Jacobian) of the constraints" in spite of providing the option to compute the Jacobian based on finite differences. Not sure if I am missing something here, but your comments will be really helpful. 

I am using the following commands: 

options.ipopt.jacobian_approximation = 'finite-difference-values';  
options.ipopt.hessian_approximation  = 'limited-memory'; 

funcs.objective         = @objective;
funcs.constraints      = @constraints;
funcs.gradient          = @gradient;

I am providing Matlab functions for objective, constraints and gradient. I tried out giving the Jacobian structure too but it does not help and the same error occurs. 

Thanks and Regards, 
Atri. 

----------------------------------------------------------------------------------- 
Atri Dutta, Ph.D. 
Post-Doctoral Research Associate, 
Mechanical and Aerospace Engineering, 
Princeton University. 

Engineering Quadrangle, Room D307, 
Princeton  NJ 08544. 
Email: atrid at princeton.edu
Homepage: www.a3dutta.net 
-----------------------------------------------------------------------------------