試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 11 / 65
A company's hiring AI is disadvantaging a particular gender. What is the most direct way to mitigate that bias via pre-processing before training, without altering the training data?
①Assign weights to each sample to adjust their influence during training.
②Completely remove protected attributes from the data.
③Automatically correct data labels to standardize them.
④Use a loss function that enforces fairness constraints during training.
⑤Increase the model's hyperparameters to improve its expressiveness.