no code implementations • 11 Jun 2019 • Brian Hentschel, Peter J. Haas, Yuanyuan Tian
To maintain the accuracy of supervised learning models in the presence of evolving data streams, we provide temporally-biased sampling schemes that weight recent data most heavily, with inclusion probabilities for a given data item decaying over time according to a specified "decay function".
no code implementations • 12 Sep 2018 • Abdul Wasay, Brian Hentschel, Yuze Liao, Sanyuan Chen, Stratos Idreos
We propose MotherNets to enable higher accuracy and practical training cost for large and diverse neural network ensembles: A MotherNet captures the structural similarity across some or all members of a deep neural network ensemble which allows us to share data movement and computation costs across these networks.
no code implementations • 29 Jan 2018 • Brian Hentschel, Peter J. Haas, Yuanyuan Tian
Moreover, time-biasing lets the models adapt to recent changes in the data while -- unlike in a sliding-window approach -- still keeping some old data to ensure robustness in the face of temporary fluctuations and periodicities in the data values.
Databases