Ask Me Anything session with a Kaggle Grandmaster Vladimir I. Iglovikov

[khuyentran1401]

TL;DR

Data science practice of high achiever in data science

Article Link

https://towardsdatascience.com/ask-me-anything-session-with-a-kaggle-grandmaster-vladimir-i-iglovikov-942ad6a06acd

Author

Vladimir Iglovikov

Key Takeaways

• Use minimal Jupyter Notebook. Write code base that can be refactored in the future. This will speed you up later on
• Write unit tests wherever possible
• Minimize the use of a mouse (write down hotkeys in a piece of paper)
• Do not use social networks
• Check email only a few times a day
• Skills in Kaggle are just the small subset of real skills in the industry
• Invest on good GPU can be a good investment over a long term
• Experiment with the parameters without the certainty why the parameters may work
• To get above the level of experimenting, you need to educate yourself with books, courses, write better code. Learn and experiment at the same time
• Read what others do and try to understand what they do. The next time you could do the same techniques with similar problems

Useful Code Snippets

Useful Tools

• PyCharm
• DVC: Version control
• Trellow: Keep track of the tasks

Comments/ Questions

• Really useful advice not just for Kaggle but for data science practice in general