DRESS

Curses and blessings of dimensionality

This project "Dimensionality Reduction for Efficient Similarity Search (DRESS)" is funded in part by EPSRC (EP/E037402/1) and Australian Research Council (DP0663272). A wide range of applications, such as multimedia retrieval, molecular biology, medical imaging, and so on, involve similarity search in high-dimensional feature spaces, which is often computational expensive if not prohibitive. Traditional index structures in relational databases and multi-dimensional index structures in spatial databases are not sufficient to index high-dimensional data. This project aims to improve the efficiency of similarity search by exploring linear and non-linear dimensionality reduction techniques and adapting them to different similarity measures, feature spaces and query processing strategies.