Algorithm

Use the radio buttons to specify an optimization algorithm.

Available options are described in the table below:

Option	Definition
Greedy Forward Selection	This process steps through the optimization, adding additional features at every step to aggressively reduce the cost function. Choose this option to achieve sparsity with the fewest possible steps. Note: This method typically runs much faster than Active Set optimization but, because it can never correct mistakes made in an earlier step, can be subject to getting “stuck” at a local optimization.
Active Set	This process identifies a reasonable starting point and then steps through the optimization, considering the effects on final optimization. It repeats each step, removing suboptimal features until an acceptable solution is reached. Note: This method typically runs much slower than Greedy Forward Selection but be the method of choice, because it can correct mistakes made in earlier steps.

Option

Definition

Greedy Forward Selection

This process steps through the optimization, adding additional features at every step to aggressively reduce the cost function. Choose this option to achieve sparsity with the fewest possible steps.

Note: This method typically runs much faster than Active Set optimization but, because it can never correct mistakes made in an earlier step, can be subject to getting “stuck” at a local optimization.

Active Set

This process identifies a reasonable starting point and then steps through the optimization, considering the effects on final optimization. It repeats each step, removing suboptimal features until an acceptable solution is reached.

Note: This method typically runs much slower than Greedy Forward Selection but be the method of choice, because it can correct mistakes made in earlier steps.

To Specify the Optimization Algorithm:

Make sure that the Optimize the continuous predictor set check box is checked.

Select the desired algorithm by clicking within the appropriate radio button.