by Jeong-Yoon Lee
Giuliano Janson gave a great perspective as an ex-Kaggle addict who successfully recovered from it. Let me give mine as an active Kaggle addict, who is helpless without any hope for rehab. ;)
I have a weekly Data Science journal club at work, which I turned into a weekly Kaggle club. I invite people from outside the company as well and now we have about 10 people joining from 4 different companies.
Today, as we did post-mortem on the Quora competiton that ended earlier this week, one of attendees, who is relatively new to Kaggle, shared that she found it very addictive.
I agreed with her and added that it is addictive because it is so to observe yourself improving, just like how work-out and career advancement can be addictive. It is a different kind of addiction from ones that harm yourself.
At the post-mortem meeting, we shared lists of things that we learned from the competition: Feature engineering in NLP with various embeddings, good ML implementations such as LightGBM, XGBoost and Keras, new deep learning architectures combining LSTM or CNN with meta features, and many more.
And we talked about which competition to enter next. Yeah, we don’t need a break. We need another dose of a competiton, or , as I like to put, a deliberate practice opportunity.
One tip for me to remain addicted for a long time - past 6 years - is that I don’t compete to rank higher, but I compete to learn more and have fun with others. That’s why I compete with a team with different people at various skill levels, including first time Kagglers.
How do I feel? I feel great to be addicted to Kaggle and even better to be addicted to self improvement. :)
At the Quora competition, we finished 36th out of 3,307 teams. One of members was a first timer. Yeah, I gave him a pretty good gateway drug. No regret! ;)