Description Usage Arguments Value Examples
Provides the default tuning parameters for gforce.FORCE.
1 |
d |
dimension of random vector or number of datapoints. |
An object with following components
adapt_init_modea numeric. Indicates which initialization mode to use for gforce.FORCE_adapt.
alphaa numeric. Gives the step size for the projected gradient descent updates.
dual_frequencyan integer. Specifies how many gradient updates to perform between searches for a dual certificate.
duality_gapa numeric. If the duality gap can be shown to be less than duality_gap, the FORCE algorithm terminates.
early_stop_modea numeric. early_stop_mode == 1 indicates that the algorithm should use an early stopping rule.
early_stop_lagan integer. This indicates the number of iterations without sufficient improvement in objective value before early stopping.
early_stop_epsa numeric. Threshold for objective value improvement used to determine early stopping.
eps_obja numeric. Specifies the precision required of the optimal solution to the eigenvalue maximization problem.
finish_pgdan integer. If finish_pgd is 1, then other stopping criteria are ignored and FORCE performs max_iter gradient updates.
initial_mixinga numeric between 0 and 1. Specifies how to construct the initial strictly feasible solution to the SDP relaxation.
kmeans_iteran integer. The number of times to run a K-means solver during each search for an optimal clustering and dual certificate.
max_iteran integer. The maximum number of gradient updates to perform.
primal_onlyan integer. primal_only == 1 indicates that the algorithm should not search for a dual certificate.
restartsa vector of integers. This specifies the iterations at which to take the projection of the current iterate and restart the algorithm with that as the initial solution.
verbosean integer. Specifies the level of verbosity requested from gforce.FORCE.
1 | opts <- gforce.defaults(20)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.