Trust Your Robots! Predictive Uncertainty Estimation of Neural Networks with Sparse Gaussian Processes

This paper presents a probabilistic framework to obtain both reliable and\nfast uncertainty estimates for predictions with Deep Neural Networks (DNNs).\nOur main contribution is a practical and principled combination of DNNs with\nsparse Gaussian Processes (GPs). We prove theoretically that DNNs can be seen\nas a special case of sparse GPs, namely mixtures of GP experts (MoE-GP), and we\ndevise a learning algorithm that brings the derived theory into practice. In\nexperiments from two different robotic tasks -- inverse dynamics of a\nmanipulator and object detection on a micro-aerial vehicle (MAV) -- we show the\neffectiveness of our approach in terms of predictive uncertainty, improved\nscalability, and run-time efficiency on a Jetson TX2. We thus argue that our\napproach can pave the way towards reliable and fast robot learning systems with\nuncertainty awareness.\n