In many problems, attempting to develop a classifier follows a typical pipeline: given labeled, raw training data, chose a classifier and hyperparmeter values to learn a predictive model. This step is repeated over and over again until the performance is satisfactory. How to choose what classifier to use or what hyperparameter value to choose? Can previous models tell us something about what could possibly be a good model to try? In this work, we develop an AutoML system which attempts to automate the classifier selection process.
T. Swearingen, W. Drevo, B. Cyphers, A. Cuesta-Infante, A. Ross and K. Veeramachaneni. "ATM: A distributed, collaborative, scalable system for automated machine learning," Proc. of 2017 IEEE International Conference on Big Data (Big Data 2017), (Boston, MA, USA), December 2017. To Appear.