ML engineering: dữ liệu, đánh giá và hợp đồng serving

Nội dung luyện tập

• ml majority baseline accuracy limit
• ml probability column class order
• ml mae original target units
• ml integer seed repeat split
• ml scaler reuses training statistics
• ml unknown category zero vector
• ml confusion matrix axis contract
• ml threshold changes decisions not scores
• ml imputer statistic from fit
• ml balanced accuracy class average
• ml preprocessing fit inside cv fold
• ml time series gap index boundary
• ml group split unseen entity evaluation
• ml roc auc ranking not calibration
• ml calibration frequency interpretation
• ml permutation correlated feature importance
• ml untrusted model artifact loading
• ml historical feature point in time
• ml training serving feature unit skew
• ml unlabeled drift not quality proof

Mỗi câu đúng 1 điểm. Mục tiêu 70%. Đáp án và giải thích mở sau khi nộp bài.