This page is part of the MAD Manual. See MAD Manual.
Conclusions & Future Work
In this document we have introduced the user to the automatic differentiation (forward mode) of Matlab expressions and functions using the fmad class. We have included examples that illustrate:
- Basic usage for expressions
- Basic usage for functions
- Accessing the internal, 2-D matrix representation of derivatives.
- Preallocation of arrays to correctly propagate derivatives
- Calculating sparse Jacobians via dynamic sparsity or compression.
- Differentiating implicitly defined functions.
- Controlling dependencies
- The use of fmad in Matlab toolboxes (ODEs and Optimization)
- Differentiating black box functions
Future releases of this document will feature reverse mode, second derivatives, FAQ's and anything else users find useful. The function coverage of fmad will be increased and present restrictions decreased with user demand. We are interested in receiving user's comments and feedback, particularly regarding cases for which fmad fails, gives incorrect results or is not competitive (in terms of run times) with finite-differences.