Selecting features using all data before splitting into folds for training/testing is a big source of train-test leakage. To demonstrate, I generated random data and labels, select down to 25 features, and train a model. Much better than random performance due to the leakage.
DeepFake:
- not as effective as you might think.
- more easily detectable as you might think.
Which may explain why trolls and propagandists haven't used it much.... https://t.co/1ZctnbhZH6
After a year of hard work, today we (pre)launch Ludwig, the A.I. assistant for accountants! Live demo at 2pm on #fastforward19@getsilverfin, c u there!
https://t.co/eaSZCl2lQ7
Boltzmann was invited to lecture at the Vlerick Business School "Big Data and Analytics for Finance and Strategy" bootcamp. During the three weeks we showed the students the theory behind machine learning and data science and its practical application.
https://t.co/sGAJ1XwSFb
Learn how to detect anomalies at Data Science Meetup Leuven, tonight. #Boltzmann's partner Tim Verdonck is giving a talk at 7.30 pm, meet us there! #meetup#datascience
https://t.co/M961nNEapD
🍾Congratulations to Boltzmann, Banking Buddy, the winners of the B-Hive #Fintech Hackathon and thank you once again to the jurors and all the teams who participated 🙌🏻