Homepage of ADMAT 2.0

The Numerical solution of large scale nonlinear problems involves computing the derivative information in the form of gradients, Jacobian and Hessian matrices.

ADMAT 2.0 enables you to differentiate MATLAB functions, and allows you to compute gradients, Jacobian matrices and Hessian matrices of nonlinear maps defined via M-files. Both forward and reverse modes are included in ADMAT 2.0. When the Jacobian and Hessian matrices are sparse, the included package ADMIT-1 provides another efficient way to compute the Jacobian and Hessian matrices. You need only supply a M-function to be differentiated and ADMIT-1 will exploit the sparsity of the present function to yield sparse derivative matrices (in sparse Matlab form). ADMIT-1 also allows for the calculation of gradients and has several other related functions. Note that ADMIT-1 is much faster than forward or reverse mode when Jacobian matrices has the same structure but need to be evaluated at different points.

ADMAT inherits the algorithms and the structures of ADMAT. In ADMAT 2.0, most "for" loops were replaced with the matrix operations in Matlab, which makes ADMAT 2.0 much faster than the old version. Also, we consider all cases when input variables are scalars, row vectors, column vectors or matrices, so that the new version will be more robust than the old one.

Download ADMAT 2.0.

Manual for ADMAT 2.0.

Algorithms and structures of ADMAT and ADMIT.

กก

If you have any questions, please feel free to contact matlab1@math.uwaterloo.ca.

กก