Elena Kasianenko
Nazar Sheremeta
14 Oct 2018
17:00 - 17:45
Stream B

Predicting Rare Events in the Car

We have built a system, which uses user location, signals from car and user behavior as input, will tell if that’s a good idea to use autonomy at this time and place, or not. This system could be used to turn on autonomy automatically, in the same way automatic gearbox switches gears in cars. This task features both personalized and fleet learning, we will talk about a challenges, trade-offs and drawbacks of each type of learning. We will describe how we usually solve this kind of tasks, while giving deep dive into Predictive Autonomy problem and techniques we used. The topic would be the most beneficial to listeners with basic ML knowledge. However, it should be accessible to everyone, since we will try to make it clear, yet interesting for everyone.