Congratulations to our PhD researcher, Gil Keren. Gil won the best student paper at the  2018 IEEE International Conference on Data Mining (ICDM), the leading research meeting in the field, presenting his work in Singapore on the 18th November.

Gil’s work faced stiff competition. A total of 948 papers were submitted to the conference, with less than one in ten submissions accepted as full papers.

Titled ‘ ‘Fast Single-Class Classification and the Principle of Logit Separation', Gil’s paper investigates the training properties of large-scale classification models. His findings demonstrate the ability of a particular family of loss functions to considerably speed up the inference of such models, which operate over a large number of classes.