Student research opportunities

Scalable Relational Kernels for Data Mining

Project Code: CECS_141

This project is available at the following levels:
Honours, Summer Scholar, Masters

Keywords:

Data Mining; Machine Learning; Kernel Methods; Relational Databases

Supervisor:

Dr Scott Sanner

Outline:

Relational kernels provide a powerful way to leverage recent advances in machine learning (kernel machines such as SVMs) with the vast data storage capabilities of databases in an efficient way. This project focuses on how to make inference with relational kernels as efficient as possible.

Goals of this project

You will design algorithms and data structures that efficiently perform kernel computations via SQL queries. When complete, you should have efficient techniques for applying any kernelized algorithm from machine learning to any SQL database.