track.control: Auxiliary for controlling tracking routines

In evanwang1990/FMwR: Factorization Machines with R

Usage

track.control(step_size = -1, evaluate.metric = "LL", convergence = 1e-04)

Arguments

step_size	record parameters every step_size iteration
convergence, evaluate.metric	control how to early stop

Details

Early stopping is controlled by evaludate.metric and convergence. If the difference of the value between two interations, which is calculated by evaluate.metric, is smaller than convergence for three times in a row, iteration will stop early. evaluate.metric includes:

• "LL":loglikehood

• "AUC": auc

• "ACC": accurancy

• "RMSE": root mean square err

• "MAE": mean absolute error

"LL","AUC","ACC" are for CLASSIFICATION task. And "RMSE","MAE" are for REGRESSION task.

evanwang1990/FMwR documentation built on May 16, 2019, 9:38 a.m.