CPC G06N 20/00 (2019.01) [G06F 17/18 (2013.01); G06N 3/084 (2013.01)] | 12 Claims |
1. A method for training a machine learning model comprising:
receiving, by a computer system comprising a processor and memory, a training data set comprising imbalanced data;
computing, by the computer system, a label density fX(x) representing a continuous distribution of labels of the training data set over a domain X as a function of independent variable x ∈ X;
computing, by the computer system, a weight function w(x) comprising a term that is inversely proportional to the label density representing a continuous distribution of labels of the training data set over a domain X as a function of independent variable x ∈ X;
weighting, by the computer system, a loss function (x, x) in accordance with the weight function to generate a weighted loss function w(x, x);
training, by the computer system, a continuous machine learning model in accordance with the training data set and the weighted loss function w(x, x) to compute a trained continuous machine learning model configured to compute a prediction of a continuous value; and
outputting, by the computer system, the trained continuous machine learning model,
where x represents a value of the training data set and where x represents a continuous value predicted by the continuous machine learning model,
wherein the weight function w(x) is computed in accordance with a weighting parameter Δ greater than or equal to 1 reflecting a ratio between a maximum weight and minimum weight of the weight function.
|