B1460
Title: Data challenges in applied AI projects: A few stories
Authors: Jean-Francois Plante - HEC Montreal (Canada) [presenting]
Abstract: Machine learning algorithms are designed to leverage structure in data in order to discover patterns, and/or to make predictions. In applied projects, the success of good ideas depends heavily on the availability of appropriate data. The crucial role of those data is typically not emphasized enough, and industrial partners may be sceptical about it. We will present a few recent examples of industrial projects where good ideas were limited by the availability or characteristics of the data. Additional challenges also come when partners get excited by generic promises that may not be applicable to their project. From sawmills to ventilation systems, we will share some key steps of three different industrial research projects in applied AI.
