While implementing the benchmark for Logistic Regression (#17), I noticed that trueml.losses currently lacks a Binary Cross Entropy loss (BCELoss) implementation. To benchmark and train binary classification models properly, we need a stable BCE loss function alongside its gradient computation.
While implementing the benchmark for Logistic Regression (#17), I noticed that
trueml.lossescurrently lacks a Binary Cross Entropy loss (BCELoss) implementation. To benchmark and train binary classification models properly, we need a stable BCE loss function alongside its gradient computation.