Hugo Lewi Hammer: Tracking of Quantiles for Dynamically Varying Data Streams
Traditional quantile estimators are not well-suited for data streams because the memory and computational time increase with the volume of data received from the stream. Incremental quantile estimators refer to a class of methods designed to maintain quantile estimates for data streams. These methods operate by making small updates to the estimate every time a new observation is received from the stream. In this presentation, I will introduce some of the incremental quantile estimators we have developed.
Hugo is a professor of statistics at the Department of Computer Science at OsloMet and also serves as an Adjunct Chief Research Scientist at SimulaMet. His primary research interests include uncertainty quantification and interpretability of prediction models, as well as extracting knowledge from them.