Alpine Meadow : A System for Interactive AutoML

AutoML has been widely used by domain experts without machine learning knowledge to extract actionable insights out of data. However, previous studies only emphasize the high accuracy of the final answer, which can take several hours, if not a couple of days to complete. In this paper we present Alpine Meadow, a first Interactive Automated Machine Learning tool. What makes our system unique is not only the focus on interactivity, but also the combined systemic and algorithmic design approach. We design novel algorithms for AutoML search and co-design the execution runtime to efficiently execute the ML workloads. We evaluate our system on 300 datasets and compare against other AutoML tools, including the current NIPS winner, as well as expert solutions. Not only is Alpine Meadow able to significantly outperform the other AutoML systems while — in contrast to the other systems — providing interactive latencies, but also outperforms in 80% of the cases expert solutions over data sets we have never seen before.