Kernel-Based Cardinality Estimation on Metric Data

On this website we provide data, queries and distance measures used in the experiments of the paper "Kernel-Based Cardinality Estimation on Metric Data".

Contact

In case of any questions or for further information contact: Michael Mattig  

Data Sets

Moby Word List (Moby)

This data set is derived from the Moby World List.

We used the Levenshtein distance measure for this data set.

Download: Data (2MB), Queries (2KB)

Protein Binding Sites (PBS)

This data set is extracted from CavBase which was introduced in
Schmitt, Stefan, Daniel Kuhn, and Gerhard Klebe. "A new method to detect related function among proteins independent of sequence and fold homology." Journal of molecular biology 323.2 (2002): 387-406.

We used the distance measure proposed in
T. Fober, M. Mernberger, G. Klebe, and E. Hüllermeier. Effcient similarity retrieval of protein binding sites basedon histogram comparison. In GCB, pages 51-59, 2010

We provide the data in histogram form as presented in this paper.

A Java implementation of this distance measure is available: Download

The Code for deserializing the binary data and queries: Download

Download: Data (521MB), Queries (186KB)

Rea16

This data set consists of 16 dimensional vectors (original source).

We used the Euclidean distance measure for these data sets.

Download: Data (177MB), Queries (14KB)

Wiki

This data set contains Wikipedia articles represented by the article name and a short abstract. We obtained the original data from the DBpedia project. We then cleaned tje data by removing articles that are just lists of links or disambiguations. We also removed all non-ASII characters.

We used a custom distance measure based on the Jaccard coefficient of character shingles of length k=3.

Download: Data (1.71GB), Queries (34KB)