The generation of high-quality training data has become the key  bottleneck in the use of deep learning across many domains. We describe  Eureka, an interactive system that leverages edge computing and early  discard to greatly improve the productivity of experts in the  construction of a labeled data set. Our experimental results show that  Eureka reduces the labeling effort needed to construct a training set by  two orders of magnitude relative to a brute-force approach.

Feng, Z.,  George, S., Harkes, J.,  Pillai, P. , Klatzky, R., Satyanarayanan, M.
Proceedings of the Third IEEE/ACM Symposium on Edge Computing (SEC 2018), Bellevue, WA, October 2018