Data Science Design Patterns
summary
Claims that most data‑science projects reduce to a few repeatable patterns and demonstrates concrete ones: adapter pattern for hiding messy Word‑doc→string→NumPy→pandas ETL, pipeline/sequential models for layered networks, strategy pattern for model selection and hyperparameter search (shows TPOT’s genetic search), plus a builder idiom discussion. Uses Python tools — NumPy, pandas, Jupyter, context managers — and recommends separating business logic, owning API design, and favoring small functional units; these practices are presented as practical and usable, not just conceptual.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:35Presentation approach
- 1:50About me
- 2:43Python
- 3:23What do they do
- 4:21Definition
- 4:58Why use them
- 6:21Game
- 9:13Pipeline
- 10:47Strategy Pattern
- 13:13Bonus Question
- 14:13Design Patterns vs Idioms
- 15:10Python Design Patterns
- 18:05Spot Quiz
- 20:34Adapter Pattern
- 23:04Questions
related talks
27mTips and tricks for data science projects with Python
29mData Science Best Practices : From Proof of Concepts to Production
46mOn Being Pythonic
33mData Science & Data Visualization in Python
37mFunctional Programming for Data Scientists
47mA Practical(ish) Introduction to Data Science
51mData Literacy
35mMental Models to Use and Avoid as a Data Scientist