aum_linear_model_ls {aum}

aum linear model ls

Description

Learn a linear model with weights that minimize AUM. Weights are initialized as a vector of zeros, then optimized using gradient descent with exact line search.

Usage

aum_linear_model_ls(feature.list, 
    diff.list, max.steps = NULL, 
    improvement.thresh = NULL, 
    maxIterations = "min.aum", 
    initial.weight.fun = NULL)

Arguments

feature.list

List with named elements subtrain and validation, each should be a scaled feature matrix.

diff.list

List with named elements subtrain and validation, each should be a data table of differences in error functions.

max.steps

positive integer: max number of steps of gradient descent with exact line search (specify either this or improvement.thresh, not both).

improvement.thresh

non-negative real number: keep doing gradient descent while the improvement in objective is greater than this number (specify either this or max.steps, not both).

maxIterations

max number of iterations of exact line search. If "max.auc" then the objective for improvement.thresh is max AUC, otherwise objective is min AUM. Default is "min.aum"

initial.weight.fun

Function for computing initial weights, default NULL means use a random standard normal vector.

Value

Linear model represented as a list of class aum_linear_model with named elements: loss is a data table of values for subtrain and optionally validation at each step, weight.vec is the final vector of weights learned via gradient descent, intercept is the value which results in minimal total error (FP+FN), learned via a linear scan over all possible values given the final weight vector, and search is a data table with one row for each step (best step size and number of iterations of line search).

Author(s)

Toby Dylan Hocking <[e-mail]> [aut, cre], Jadon Fowler [aut] (Contributed exact line search C++ code)