SCS Undergraduate Thesis Topics
|Haijie Gu||John Lafferty||An Efficient Algorithm for Nonparametric Online Prediction|
Online learning processes input data as a stream, predicting a label for each incoming instance based on the observations in the history. Most of the current approaches to online learning focus on a single learning task. In this senior thesis research, we will study problems of multiple parallel online predictions under computation constraints. Our goal is to find approaches to model the prediction loss and cost and also, by carefully allocating the resources, to minimize the overall loss across tasks with fixed total cost.