Home

Welcome to the AD_Deriv wiki!

This wiki provides details on how to use Deriv, a suite of automatic differentiation tools in Matlab based on operator overloading.

Overview

1. Summary of the Deriv functions
2. Command-line examples
3. Calling Deriv inside functions
4. Implementation Details

Summary of the Deriv functions

• Deriv.m

The main Deriv class. This file contains the Deriv class definition (every double variable has associated with it the derivative of that variable with respect to the parameter of interest).

• Dzeros.m

A work around. If a matrix is created as a double, then an entry is defined with a Deriv object, the matrix won't correctly get promoted to a Deriv object. This function should be used instead of zeros when preallocating storage in Matlab. Use like zeros.

• Get_gradient.m

A function that computes the gradient of a scalar function by repeated evaluation and forward mode automatic differentiation. The plan is to replace this with one reverse mode call.

• Get_jacobian.m

Similar to Get_gradient, but for vector valued functions. If the nonzero entries of the Jacobian are known, this function exploits this using a sparse Jacobian strategy to reduce the number of function evaluations.

• LICENSE.md

The GPL license.

• README.md

A modest overview of the Deriv class.

• Revers.m

(empty) Will ultimately contain the machinery to implement reverse mode of automatic differentiation.

• Set_variable.m

A user friendly means to define the independent variables for which functions will be differentiated with respect to.

• test_Deriv.m

A number of unit tests of Deriv operator overloading. This can be used as an example of how to use Deriv.