L2CVSearch

This is a function for performing cross validation for optimizing the L2 regularization parameter. It uses a basic hill climbing algorithm to search a range of parameters. It returns: cv_likelihoods, a matrix of the cross validated log likelihoods for each fold; L2_reg_params, the log L2 parameters tested and mean_cvpl, the mean log likelihood for each L2 parameter tested.

Parameters:

• x_train - training set matrix. Expected numpy array.
• ytime_train - time of death or censoring for each patient
• ystatus_train - censoring of each patient
• model_params = {} - model parameters, see trainCoxMlp
• search_params = {} - model search/training parameters, see trainCoxMlp
• cv_params - cross validation parameters, see below.
• verbose=True - print more stuff.

cv_params is a dictionary of parameters for cross validation. It has the following parameters:

• cv_seed - Random seed for splitting validation folds. Default is 1.
• n_folds - Number of folds to cross-validate. Default is 10.
• cv_metric - Performance metric for evaluating cross-validation performance. loglikelihood or cindex. Default is loglikelihood.
• search_iters - Number of iterations in hill climbing. Default is 3.
• L2_range - Range to search for L2 parameter. Default [-5,-1].