Structuring Machine Learning Projects

在 Structuring Machine Learning Projects 中

將傳授一些在別的地方不會說的 experiences

這些都是來自吳恩達的 "industry experience"

• 知道如何 diagnose errors in maching learning system

• 能夠優先找到正確的 implement directions 並且減少錯誤

• 了解複雜的 ML settings

  • mismatched training / test sets

  • comparing to and/or surpassing humal-level performance

• 學會 End-to-end learning, transfer learning, multi-task learning

他見過太多 teams 浪費數月、數年在進行專案，只因為沒搞懂以上這些事情

所以這堂課可以 save 你非常多時間

這將是 Deep learning 的第三堂課 !

ML Strategy (1)

• Orthogonalization

• Single number evaluation metric

• Train/dev/test distributions

• human-level performance

• Avoidable bias

• model performance

ML Strategy (2)

• error analysis

• clean up incorrect labeled data

• Build your first system quickly, then iterate

• Training and testing on different distributions

• Bias and Variance with mismatched data distributions

• Addressing data mismatch

• Transfer learning

• Multi-task learning

• End-to-end deep learning