[HKML] Hong Kong Machine Learning Meetup Season 1 Episode 6


  • Wednesday, January 23, 2019 from 7:00 PM to 8:00 PM



In his talk, Jules presented us the world of real-time bidding (RTB): advertisers compete in an auction by bidding for an advertising space made available by a publisher; the advertiser that wins the auction is allowed to publish its content (which is usually tailored to the user by leveraging browsing history data, etc.). Some firms, like Criteo, specialize in these auctions. These firms get paid by their customers only if the user clicks on the ad; when they think there is a high probability of user click, they participate in the auction to win the advertised space, in the meantime they customize the advertisement that will be displayed to the targeted user, and bid an appropriate amount so that winning the auction makes sense economically for them. There is a lot of machine learning in this space: from estimating click probability to ad customization and A/B testing. Jules explained how feature engineering is important yet difficult, and why logistic regression (a relatively simplistic model) is (was?) the preferred model in this industry (super fast to train/predict with respect to other models). Details in his slides.

Félix, based on his data science consulting experience in big corporates, delivered a presentation on data science project management: How to make an effective impact using machine learning in the context of big non-tech companies, in 10 steps. Here, his slides, if you want to know more!