Bobyqa

This block finds the least value of a function of many variables by applying a trust region method that forms quadratic models by interpolation. Some freedom is usually present in the interpolation conditions, which are taken up by minimizing the Frobenius norm of the change to the second derivative of the model, beginning with the zero matrix. The values of the variables are constrained by upper and lower bounds.

    Bobyqa

Library

Activate/Optimization

Description

This block implements the Bobyqa optimization program in reverse communication mode. When activated, the block expects to receive on its input ports respectively the value of the optimization vector p and the associated cost. The block then produces on its first output either a new guess value of p for which it expects a cost evaluation at the next iteration, or the optimal solution. In the latter case, the block sets its output, iter (second output), to zero, and the last output, cost_opt, to the cost of the optimal solution. The output iter indicates whether or not an additional iteration is needed. If iter=0, then the optimal value of the cost associated with the optimal p is also produced on the last block output.

For more details, see the section on Optimization in the Extended Definitions for Advanced Users.
For further explanation of the BOBYQA algorithm, see: www.damtp.cam.ac.uk/user/na/NA_papers/NA2009_06.pdf

Parameters

Bobyqa_0

NameLabelDescriptionData TypeValid Values

xL

Variables lower bounds

Matrix of size Mx1

xU

Variables upper bounds

Matrix of size Mx1

RhoBeg

Initial value of the trust region

Matrix of size 1x1

RhoEnd

Final value of the trust region

Matrix of size 1x1

Npt

Number of interpolation conditions

Scalar

MaxIter

Maximum number of iterations

Scalar

Ports

NameTypeDescriptionIO TypeNumber

p0

explicit

input

1

cost

explicit

input

2

p

explicit

output

1

iter

explicit

output

2

cost_opt

explicit

output

3

Port 6

activation

input

1

See Also