In machine learning, bias refers to systematic skew or distortion in training data. It arises when certain groups or situations are over- or under-represented, or when data sources and creation histories are unbalanced. Models trained on biased data risk producing unfair or inaccurate predictions, making mitigation measures — such as data quality checks, provenance verification, and consistent annotation standards — essential.

Past questions testing this term · 1 questions

• 2025r07 #8AIの機械学習で利用するデータの取扱いに関する記述のうち，バイアスの低減やデータの品質を確保するために考えられる対策として，適切なものだけを全て挙げたものはどれ…