Mediation Modeling at Uber: Understanding Why Product Changes Work (And Don't Work)
By: Bonnie Li
At Uber, we apply mediation modeling, a statistical approach from academic research, to address user pain points. Mediation modeling goes beyond simple cause and effect relationships in an attempt to understand what underlying mechanisms lead to a given result. Using this type of analysis, we can fine-tune product changes and develop new ones that focus on the underlying mechanisms behind successful features on the Uber platform.
An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solution
By: Rosanne Liu
Convolution is prevalent in deep learning. However, recent work from Uber AI unfolds one of its generic inabilities in transforming spatial representations between different types of coordinate systems. Such a finding as well as a subsequently revealed solution in the form of the CoordConv layer have proved to bring about improvements in a wide range of domains, including object detection, generative models, and reinforcement learning.
Food Discovery with Uber Eats: Recommending for the Marketplace
By: Yuyan Wang
The Uber Eats marketplace consists of three sides: eaters, restaurant-partners and delivery-partners. Ranking and recommending restaurants on Uber Eats is therefore a unique challenge, as each each side has intrinsic and different values that we need to take into consideration when optimizing for the marketplace. In this talk, we will go through the journey of ranking restaurants on the Uber Eats app, and zoom into the multi-objective optimization framework for our three-sided marketplace.