User Tools

Site Tools


Data Science Seminar
Hosted by Department of Mathematical Sciences

  • Date: Tuesday, March 13, 2018
  • Time: 12:05pm – 1:05pm
  • Room: WH-100E
  • Speaker: Jiexin Duan (Purdue University)
  • Title: Large-Scale Nearest Neighbor Classification with Statistical Guarantee


In this talk, we develop distributed classification methods based on nearest neighbor principle. Through majority voting, the distributed classification can achieve the oracle rate of regret and instability if neighbor neighbors are carefully chosen. The only loss is a multiplicative constant that depends only on data dimensionality. This can be remedied by replacing majority voting with continuous aggregation.

seminars/datasci/180313.txt · Last modified: 2018/02/19 21:57 by qiao