試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 1 / 65
What is the problem called when the distribution of input features changes between training and deployment, and what countermeasures are effective?
①It's called overfitting and can be solved by data augmentation alone.
②This is called underfitting and can be solved by increasing the model's capacity.
③Adjusting the temperature parameter, known as label smoothing, is effective.
④It's a covariate shift (distribution shift) and can be mitigated by domain adaptation or online retraining.
⑤This is called data leakage and can be resolved by increasing the number of features.